AI Crypto Agents Are Ushering in a New Era of ‘DeFAI’

Imagine your investments working around the clock, scanning global markets for the best opportunities — all without you having to lift a finger. Sound futuristic? It’s already a reality.

In traditional finance (TradFi), algorithms handle nearly 70% of U.S. stock trades. Now, artificial intelligence (AI) agents are stepping up. These aren't just basic bots but innovative systems that learn, adapt and make real-time decisions. VanEck predicts the number of AI agents will skyrocket from 10,000 to over a million by the end of 2025.

What this means for you

AI agents are already at work behind the scenes analyzing market trends, balancing portfolios and even managing liquidity across decentralized exchange platforms like SaucerSwap and Uniswap. They're blurring the lines between TradFi and decentralized finance (DeFi), with cross-chain transactions expected to jump 20% in 2025.

Can we really trust AI with our money?

Autonomous finance isn’t new, but today’s AI agents operate with increased autonomy and sophistication. So, can we trust these agents to manage billions in digital assets? What safeguards exist when decisions come from algorithms, not humans? Who would be held responsible for market manipulation performed by an agent?

These concerns are valid. As AI agents take on more responsibility, and especially as the convergence between crypto and TradFi accelerates, worries around transparency and market manipulation will grow. For example, some blockchains enable front running trades and sandwich attacks that can exploit blockchain consensus in a process known as Maximal Extractable Value (MEV). These transaction strategies harm fairness and market trust. Operating at machine speed, AI agents could supercharge these risks.

Enter DLT: the trust layer we need

Trust is key, and distributed ledger technology (DLT) offers a solution. DLT provides real-time transparency, immutability and decentralized consensus, ensuring decisions are trackable and auditable. The Identity Management Institute reported companies that integrated blockchain identity systems have already cut fraud by 40% and identity theft by 50%. Applying these guardrails to AI-driven finance can counter manipulation and promote fairness. Moreover, the use of DLTs with fair ordering is growing rapidly, ensuring transactions are sequenced fairly and unpredictably, addressing MEV concerns and promoting trust in decentralized systems.

DeFAI: where finance is headed

A blockchain-powered, trust-centric model could unlock a new paradigm, “DeFAI”, in which autonomous agents can operate freely without sacrificing oversight. Open-source protocols like ElizaOS, which have blockchain plugins, are already enabling secure and compliant AI interactions between agents across DeFi ecosystems.

Bottom line: trust will define the future of AI

As AI agents take on more complex roles, verifiable trust becomes non-negotiable. Verifiable compute solutions are already being built by firms like EQTY Lab, Intel and Nvidia to anchor trust on-chain. DLT ensures transparency, accountability and traceability. This is already in motion; on-chain agents are now operating that offer services ranging from trade execution to predictive analytics. We can trust AI when we have trust in the model input and output.

The question now isn’t if institutions will adopt autonomous finance, but whether frameworks can evolve fast enough. For this revolution to thrive, trust must be embedded into the foundation of the system.

Floki Teams With Softbank Partner Rice Robotics for Tokenization of AI Data

Rice Robotics and dog-themed utility project Floki will soon launch the Minibot M1, a Floki-branded AI-powered companion robot that works on the RICE AI platform.

Floki will support the company in its blockchain push as it moves towards tokenizing its brand and AI data marketplace through TokenFi, a Floki sister project focused on tokenizing real-world assets, with support from the Floki community.

“The AI robotics market size is currently worth an estimated $22 billion and is projected to reach $100 billion by 2030, and we believe Rice Robotics is well-positioned for growth in this high-potential industry,” the Floki team told CoinDesk in a Telegram message.

RICE AI is a robotics brand with high-profile clients such as Nvidia, Softbank, Dubai Future Foundation, Mitsui Fudosan, NTT Japan, and 7-Eleven. It raised over $7 million in a pre-Series A funding round earlier this year from investors including Alibaba Entrepreneurs Fund, Soul Capital and Audacy Ventures.

RICE AI wants to make robots smarter by creating a system where robots worldwide can buy and share top-notch training data. These robots work independently without central control, making them more useful in the real world.

FLOKI prices are up 16% in the past 24 hours alongside a broader crypto market bump.

Galaxy , CoreWeave Expand Alliance With Data Center Expansion

Galaxy Digital (GLXY) said it deepened its strategic partnership with CoreWeave (CRWV), reinforcing its ambitions in the fast-growing artificial intelligence (AI) and high-performance computing (HPC) data center industry.

Under a new agreement, CoreWeave will gain access to an additional 260 megawatts (MW) of critical IT load at Galaxy’s Helios data center campus in West Texas, bringing the total committed capacity for AI and HPC operations at the site to 393 MW.

The move marks another shift by Galaxy away from bitcoin mining, with the Helios campus acquired from Argo Blockchain in 2022 moving toward becoming a cornerstone for next-gen digital infrastructure. CEO Mike Novogratz emphasized the strategic value of diversifying the company's business across blockchain, crypto and AI, highlighting the long-term potential to maximize shareholder value.

Galaxy shares rose as much as 8% in Toronto trading and are now up 60% from their April lows. CoreWeave rose as much as 13% as was recently trading 10% higher.

This announcement follows the March Phase I lease agreement that covered 133 MW over 15 years. The new Phase II commitment mirrors the terms of the initial deal and reflects both parties' confidence in the site's capacity and strategic location. With infrastructure upgrades already in motion, Phase I is expected to be service-ready by mid-2026, while Phase II will come online in 2027.

The site benefits from 800 MW of approved capacity and an additional 1.7 gigawatts currently undergoing evaluation — positioning Galaxy for further expansion.

CoreWeave retains exclusivity for even more capacity

Meanwhile, Galaxy is also exploring opportunities to monetize its legacy bitcoin mining infrastructure, signaling a decisive pivot in its operational focus.

Disclaimer: This article, or parts of it, was generated with assistance from AI tools and reviewed by our editorial team to ensure accuracy and adherence to our standards. For more information, see CoinDesk’s full AI Policy.

What Is TAO, the Bittensor Token Causing Friction Between Barry Silbert and Bitcoiners?

TAO, the native token of AI-focused blockchain Bittensor, has been causing tension on X between Digital Currency Group founder Barry Silbert and staunch supporters of bitcoin (BTC), the original and largest cryptocurrency.

Author and bitcoin supporter Parker Lewis called Silbert and Raoul Pal, the presenter of The Journey Man podcast, a group of “affinity scammers” for promoting TAO on a recent episode.

Silbert responded by writing: “calling $TAO a scam is such a lazy attack. do better”

Grayscale Investments, one of Digital Currency Group's subsidiaries, runs a Bittensor Trust that currently has around $8 million in assets under management. It also has a spot bitcoin exchange-traded fund (GBTC) with $16.6 billion under management as well as a bitcoin mini trust ETF.

What is TAO?

Silbert sparked the bitcoiners' ire by comparing Bittensor to the Bitcoin blockchain.

“It's just like bitcoin, there was a white paper that turned into code then launched and it has the same token economics,” he said in the podcast.

While there are some similarities to BTC in that TAO's supply is capped at 21 million tokens and it goes through block reward halving events, there are also stark differences in terms of the project's ethos and use case.

Bittensor is a decentralized network that merges blockchain technology with machine learning. It was designed to become a peer-to-peer AI market, where users can share and monetize AI models.

Bitcoin was spawned out of the Libertarian cypherpunk era and designed primarily as a peer-to-peer payment method that avoided government-issued currency. In recent years has also emerged a store of value, becoming a mainstay on company balance sheets to mitigate rising inflation.

The TAO token was released two years ago and has experienced extreme volatility, rising to above $700 on two occasions in 2024 before cratering to around $200 both times. It was trading recently around $339.

Bitcoin, meanwhile, has risen from $22,000 since the start of 2023 to as high as $109,000 in January. While it's had its ups and downs, they're not as marked as the plunges typically seen across the altcoin market. Bitcoin is currently priced around $90,000 with a market cap about $1.8 trillion. TAO has a market cap around $2.98 billion, according to data on CoinMarketCap.

Auradine Raises $153M Series C for Bitcoin Mining, AI Data Center Networking

Auradine, a maker of computing equipment for bitcoin (BTC) mining and AI applications, said it raised $153 million in a Series C funding round.

The Silicon Valley, California-based company also formed a new business group, AuraLinks AI, focused on open-standards to address cooling requirements of next-generation AI data centers.

AI data centers and BTC mining share similarities in their operational requirements. Given the proliferation of AI in mainstream use in recent years, the subject of data centers is now commonplace in public discourse. This is significant for the cryptocurrency industry because most things that relate to AI data centers could also be applied to bitcoin mining.

“Our dual focus on Bitcoin and AI infrastructure places Auradine at the intersection of pivotal technologies that will reshape computing and energy utilization for decades to come,” CEO Rajiv Khemani said in a statement.

The funding round, which took Auradine’s total backing to $300 million, was led by StepStone Group and included another contribution from bitcoin miner MARA, as well as Maverick Silicon, Samsung Catalyst Fund and Qualcomm Ventures, among others.

The Protocol: Nvidia To Manufacture AI Supercomputers in U.S., New Opportunities for Crypto Miners

Welcome to The Protocol, CoinDesk’s weekly wrap-up of the most important stories in cryptocurrency tech development. We’re Margaux Nijkerk and Sam Kessler, reporters on CoinDesk’s Tech team.

In this issue:

Can Ethereum Be Truly Private? Developers Push for Encrypted Mempool, Default Privacy

Nvidia Moves AI Supercomputer Production to U.S., Opening New Avenues for Crypto Miners

MIT-Incubated Optimum Raises $11M Seed Round to Build Web3’s Missing Memory Layer

Noble’s New ‘AppLayer’ Lets Developers Build Stablecoin Tools on Celestia

This article is featured in the latest issue of The Protocol, our weekly newsletter exploring the tech behind crypto, one block at a time. Sign up here to get it in your inbox every Wednesday.

Network News

PRIVACY HEATS UP AMONG ETHEREUM DEVS: When the U.S. government sanctioned the Ethereum-based crypto mixing service Tornado Cash in 2022, it ignited a debate within the crypto community that continues three years later. Advocates argued that complying with the sanctions amounted to censorship — undermining a fundamental cypherpunk principle. President Donald Trump supported the cypherpunks and lifted the sanctions on Tornado Cash in March of this year, but for some Ethereum developers, the situation highlighted a flaw within the network that still exists today: Why should users depend on third-party apps to transact privately on the network? Perhaps emboldened by the recent Tornado Cash developments, Ethereum developers and researchers have once again begun discussing ideas for making the Ethereum network private at its core. “Privacy must not be an optional feature that users must consciously enable — it must be the default state of the network,” said PCaversaccio, whose post outlined his vision for a privacy-oriented Ethereum roadmap. “Ethereum’s architecture must be designed to ensure that users are private by default, not by exception.” In response to PCaversaccio’s post, Ethereum co-founder Vitalik Buterin left a comment on the network’s main developer forum with his own much shorter privacy-oriented Ethereum roadmap. Buterin suggested focusing on privacy for on-chain payments, anonymizing on-chain activity within applications, making communication on the network anonymous, and privatizing on-chain reads. To achieve all of this, Buterin listed various steps like integrating certain third-party privacy features into the core network. — Margaux Nijkerk and Sam Kessler Read more.

NVIDIA AI SUPERCOMPUTER PRODUCTION PLANS COULD BENEFIT CRYPTO MINERS: Nvidia plans to manufacture its next generation of AI chips and supercomputers entirely in the U.S. for the first time, the company said in a statement. The move reflects rising demand for AI infrastructure and a broader push to localize advanced tech manufacturing — one that could also benefit crypto miners repurposing their facilities for AI and high-performance computing (HPC). Many of these operators already have access to the large-scale power and cooling systems needed for data center operations, making them potential players in the growing AI economy. Crypto miners, once singularly focused on hashing power, are increasingly looking for ways to fit into the AI and HPC supply chain. Their existing access to power-dense infrastructure and logistical experience in running industrial-scale operations gives them a foothold as demand for AI computation surges. Recent tariffs by U.S. President Donald Trump, however, is causing anxiety among miners as the policy changes are expected to raise costs on ASIC miners, electrical components, networking hardware and more.— Helene Braun Read more.

MEMORY LAYER OPTIMUM RAISES $11M IN SEED: Optimum, a decentralized, performance-enhancing memory layer for any blockchain, raised an $11 million seed round, inviting its creators from institutions like Harvard and MIT to jump from the world of academia into the commercial crypto arena. The seed round was led by 1kx with participation from Robot Ventures, Finality Capital, Spartan, CMT Digital, SNZ, Triton Capital, Big Brain, CMS, Longhash, NGC, Animoca, GSR, Caladan, Reforge and others. ​​Optimum is building what it calls the missing memory layer of blockchains, making the way data is stored, accessed and propagated, faster, cheaper and truly decentralized, according to a press release. At the core of Optimum’s innovation is a method of decentralized coding for distributed systems, known as Random Linear Network Coding (RLNC), developed by Muriel Médard, an MIT professor. — Ian Allison Read more.

NOBLE’S NEW ‘APPLAYER’ LETS DEVELOPERS BUILD STABLECOIN APPS ON TOP OF CELESTIA: Noble, a blockchain for issuing real-world assets (RWA) and stablecoins, announced Wednesday that it will expand its platform by introducing “AppLayer,” an Ethereum-compatible rollup that allows developers to create their own RWA applications and infrastructure. Noble’s AppLayer aims to let developers build new financial tools optimized for real-world assets like stablecoins — digital assets whose value is pegged to another asset, like the U.S. dollar. AppLayer will leverage Celestia, a data availability blockchain that aims to bring down storage costs for data-intensive blockchain networks. Celestia, like Noble, is plugged into the Cosmos blockchain ecosystem and is compatible with the Ethereum Virtual Machine (EVM), meaning it can read smart contracts from other Ethereum-based chains. — Margaux Nijkerk Read more.

In Other News

Mantra’s OM token fell from over $6 to under $0.45 in a matter of hours on Tuesday with no apparent catalyst. CEO John Mullin said in an X post on Wednesday that he would burn his team’s tokens to win back the trust of the Mantra community. Mullin said the price drop resulted from exchanges closing OM positions, but members of the crypto community cast blame on the Mantra team. OKX founder Start Xu referred to the incident as “a big scandal.” — Jamie Crawley Read more.

Aiming to perhaps replicate Strategy’s bitcoin (BTC) playbook, except with solana (SOL), fintech commercial real estate platform Janover (JNVR) has built a SOL stack worth roughly $21 million and seen its share price rise nearly 20-fold in less than a month. The company purchased earlier this week another 80,567 SOL tokens valued at approximately $10.5 million, bringing its total holdings to 163,651. — Krisztian Sandor Read more.

DWF Labs is investing $25 million in World Liberty Financial (WLFI), the decentralized finance protocol backed by U.S. President Donald Trump and his family. The crypto market maker is also entering the U.S. market with a new office in New York City as part of its broader expansion plans, according to a press release. — Francisco Rodrigues Read more.

Regulatory and Policy

The Securities and Exchange Commission (SEC) is not yet ready to make a decision on two critical features that issuers of the spot crypto exchange-traded funds (ETFs) are hoping to add to their products. The regulator delayed a decision on whether it will allow in-kind redemptions for WisdomTree’s Bitcoin Fund (BTCW) and VanEck’s Bitcoin Fund (BITB) and Ethereum Fund (ETHW). It also moved its deadline for a decision in regards to a proposal by Grayscale to allow staking its Ethereum Trust (ETHE) and Mini Ethereum Trust (ETH), which the asset manager’s exchange, NYSE Arca had requested in February. — Helene Braun Read more.

Seychelles-based cryptocurrency exchange OKX is expanding to the U.S. and establishing a new regional headquarters in San Jose, California. The exchange will rolling out access to its platform and its native OKX Wallet to U.S.-based crypto traders.— Cheyenne Ligon Read more.

Search giant Google will only allow cryptocurrency exchanges and software wallets to advertise in the European Union if they hold a license under the EU’s Markets in Crypto-Assets (MiCA) regulation, starting April 23, the company announced. Google said advertisers must now obtain a certification from the company and demonstrate they are registered as a Crypto-Asset Service Provider (CASP) under MiCA. The company also requires advertisers to comply with any additional country-specific legal obligations.—Francisco Rodrigues Read more.

Calendar

April 30-May 1: Token 2049, Dubai

May 14-16: Consensus, Toronto

May 19-23: Solana Accelerate, New York City

May 20-22: Avalanche Summit, London

May 27-29: Bitcoin 2025, Las Vegas

June 30-July 3: EthCC, Cannes

Oct. 1-2: Token2049, Singapore

AI Crypto Tokens Nurse Losses as Nvidia Bearish Options Bets Cross the Tape

Tokens associated with artificial intelligence (AI) fared worse than the biggest cryptocurrencies over the past 24 hours. The relative weakness comes amid unusual activity in put options tied to shares of Nvidia (NVDA), the chipmaker that on Monday said it will will start building its AI supercomputers in the U.S.

While bitcoin (BTC), the largest cryptocurrency by market value, added 0.6% over 24 hours to $85,500, TAO, the token of blockchain-based machine learning network Bittensor, traded 3.6% lower at $239 and decentralized GPU rendering platform Render Network’s RNDR token was 1.7% down at $3.93, according to data source Coingecko. Other tokens, including FET, SEI and GRT lost 2%.

Nvidia short-dated put options saw notable activity on Monday, according to data tracked by Convex Value. The action was concentrated in the $100 strike put options expiring on April 17, April 25 and May 2. Additionally, there was activity in the $60 put expiring on April 17 and $50 and $85 strike puts expiring on May 16.

Convex Value called the activity in these so-called out-of-the-money put options at strikes below the Santa Clara, California-based company’s spot price of $110 unusual. “My bet would be [these are] protective plays,” an analyst at the platform told CoinDesk.

Buying a put option is akin to buying insurance against market slides. Traders typically snap them up when looking to profit from or hedge their spot/futures bets from a potential market decline.

“Someone knows something,” Substack-based analytics service Merlin Capital posted on X.

Riot Platforms Hits Post-Halving Bitcoin Production High as It Expands AI Capacity

Riot Platforms (RIOT) reported strong operational performance in March 2025, highlighted by continued expansion into the artificial intelligence (AI) and high-performance computing (HPC) sector.

The company’s bitcoin (BTC) production last month rose to 533 BTC, the most since the reward halving almost a year ago. The figure represents a month-on-month increase of 13% and 25% more than a year before. Bitcoin holdings grew to 19,223 BTC.

Riot said it plans to “aggressively pursue” development of its Corsicana facility to capitalize on rising demand for compute infrastructure used in AI and HPC.

A recently completed feasibility study by industry consultant Altman Solon confirmed the significant potential of the site to support up to 600 megawatts of additional capacity for AI/HPC applications. Key advantages include 1.0 gigawatt of secured power, 400 MW of which is already operational, 265 acres of land with substantial development potential and close proximity to Dallas — a major hub for AI and cloud computing.

The study noted the site’s ability to support both inference and cloud-based workloads, strengthening its appeal to AI/HPC tenants.

Riot maintained a steady deployed hash rate of 33.7 EH/s, while its average operating hash rate grew 3% month-over-month to 30.3 EH/s—representing a 254% increase year-over-year. Although power credits declined due to seasonal factors, Riot kept its all-in power cost low at 3.8 cents per kWh, and improved fleet efficiency to 21.0 J/TH, a 22% improvement from the previous year.

Riot’s shares fell 5.5% Friday, while the Nasdaq 100 index dropped 2.8%. They have lost 35% year-to-date.

Disclaimer: This article was generated with AI tools and reviewed by our editorial team to ensure accuracy and adherence to our standards. For more information, see CoinDesk’s full AI Policy. This article may include information from external sources, which are listed below when applicable.

OpenAI’s $40B Raise Calms Market Jitters, Sends CoreWeave and AI Tokens Higher

CoreWeave (CRWV) shares rose more than 38% on their third day of trading debut after raising nearly $1.5 billion from its IPO following OpenAi’s announcement of a record-breaking $40 billion funding round on Monday.

The artificial intelligence (AI) startup went public on the Nasdaq exchange on Friday afternoon. The stock dropped below its IPO price to $39 and ended the day flat at $40 before dropping another 10% on Monday, its first full day of trading. CoreWeave’s IPO came at a time of strong anxiety and uncertainty in global markets, diminishing investor appetite and risk tolerance.

However, investor sentiment towards AI-related stocks seemed to have changed on Tuesday after AI powerhouse OpenAI announced on Monday that it had closed a $40 billion funding round, valuing the company at $300 billion. The move appears to have assured investors that there continues to be a strong appetite for AI companies, even in the current rough market.

This positive outlook has also spread to digital assets, as AI-related tokens were boosted on Tuesday. AI tokens, including Near Protocol (NEAR), Internet Computer (ICP), Bittensor (TAO) and Render (RENDER), were all up over 3% on Tuesday, with RENDER leading the group, trading 7.4% higher. The broader digital assets market, CoinDesk20 Index, also rose 3%.

Meanwhile, the shares of Core Scientific (CORZ), the bitcoin miner and data center with a large partnership with CoreWeave, have also risen more than 9% on Tuesday.

IREN Calling Off Bitcoin Mining Expansion in Favor of AI Data Centers

Australia-based bitcoin miner IREN is redirecting its growth plans away from BTC mining and towards its AI data centers and AI cloud services businesses.

“As we near completion of our 50 EH/s mining expansion, our focus is shifting to the next phase of growth and delivering scalable infrastructure for AI and HPC,” said co-founder and CEO Daniel Roberts in a Monday business update.

Once completed at 52 EH/s, the mining expansion is expected to generate $528 million in annual cash flow, according to the company. Current installed capacity is 35 EH/s and completion is expected in the coming months.

IREN shares are lower by 2.1% premarket.

Possible Blow to Crypto as CoreWeave Reportedly Slashes Valuation to $23B

CoreWeave is looking to downsize its initial public offering just one day before hitting the market, Semafor reported.

The AI infrastructure firm was previously expected to raise $3 billion at a $30 billion valuation, according to the story, but the size has been cut and the valuation lowered to just $23 billion.

A separate story from Bloomberg said CoreWeave is now looking to raise only $1.5 billion.

CoreWeave is in close partnership with bitcoin miner Core Scientific (CORZ), which was expected to profit from the IPO if the results are positive and sustain strong revenue growth over the next few years.

In early U.S. trading, CORZ is up fractionally, but down sharply over the past month and for 2025 as a whole. AI-related tokens NEAR, ICP, RENDER have added modestly to earlier losses.

CoreWeave saw $1.9 billion in revenue in 2024 amid surging demand for AI services. However, some believe that CoreWeave’s new $12 billion deal with AI giant OpenAI could have bigger implications for the company than its IPO plan.

CoreWeave’s pullpack comes as tech stocks have lagged other market sectors since the start of this year, partially as a result of on-and-off tariffs imposed by U.S. President Donald Trump and concerns about the spending of AI companies.

CoreWeave will debut on the Nasdaq on Friday, becoming the first AI company to hit the stock market. A representative from the company could not be reached for comment at time of publication.

Michael Saylor’s $200 Trillion Bitcoin Strategy: U.S. BTC Domination and Immortality

It’s the year 2045. Digital assets move at the speed of light. AI agents interact millions of times a second, using bitcoin as a base currency. Bitcoin is now a $200 trillion asset class, a settlement layer for the AI Age of the Internet.

This is the future imagined by bitcoin evangelist Michael Saylor, the executive chairman of Strategy (MSTR). Saylor pioneered the bitcoin corporate treasury – turning his flailing software firm into a Nasdaq-listed $85 billion leveraged bitcoin powerhouse.

CoinDesk recently sat down with Saylor, Bitcoin’s ultimate maximalist, for a two-hour interview to break down his vision for global bitcoin domination.

Since the election of U.S. President Donald Trump, bitcoin has maintained a 26% gain, peaking at a $2.1 trillion market cap, and touching a January all time high of $109,000. Strategy, a Wall Street proxy for bitcoin, remains strong with about a 50% gain, despite dropping approximately 30% from November highs amid a broader decline in U.S. equities, the U.S. 10-year Treasury yield, and oil.

The United States went from regulating crypto by enforcement and covertly de-banking digital asset firms, dubbed “Operation Chokepoint 2.0” by the industry, to declaring that the U.S. will become a bitcoin superpower and the crypto capital of the world. For Saylor, the sea change means doors that were previously closed are opening. Governments and traditional institutional investors around the world that used to be afraid of engaging with digital assets are now curious.

Saylor said he is fielding invitations to speak at all the elite gatherings: South America’s 100 wealthiest families, Middle Eastern sovereign wealth funds, Morgan Stanley’s prestigious tech conference, CPAC, and the White House. He has gone from encouraging corporations to adopt bitcoin treasuries to advising nation states on establishing strategic bitcoin reserves.

Bitcoin has reached “escape velocity,” he said, because once the U.S. government begins to acquire it aggressively, the U.S. will become a beneficiary and force every country to adopt bitcoin as the global capital.

“It becomes a fait accompli,” said Saylor. “It’s one of those geopolitical moves that when you embrace the network, you force all of your allies first to adopt it, and then all your enemies have to adopt it.”

U.S. Bitcoin Strategic Reserve

President Trump’s executive order to establish a U.S. Bitcoin Strategic Reserve represents a milestone in realizing bitcoin’s manifest destiny. At one point, the U.S. held about 400,000 bitcoins, but sold half of it for proceeds of $366 million. Trump’s crypto czar David Sacks lamented that the cost to American taxpayers for selling this bitcoin prematurely is $17 billion at current market value.

The executive order directs the Secretary of the Treasury to never sell the United States’ bitcoin and to develop budget neutral ways to acquire more bitcoin. It further directs the creation of a digital asset stockpile, a portfolio of seized crypto assets that can be managed and rebalanced as necessary.

At President Trump’s White House Digital Assets Summit on March 7, Saylor proposed that the U.S. acquire 5%-25% of the total bitcoin supply by 2035 that could generate an estimated $100 trillion in economic value by 2045.

When asked about this proposal, Bo Hines, Executive Director of the Presidential Council of Advisers for Digital Assets, told CoinDesk the Trump administration wants the U.S. to acquire as much bitcoin “as we can possibly get” and is considering various creative methods, including Senator Cynthia Lummis’ (R-Wyo) proposal to use Federal reserve earnings and gold certificates to buy bitcoin.

As the U.S. embraces bitcoin, worldwide banks will inevitably follow.

“ Pandora’s box has been opened,” said Saylor. “When bitcoin spreads… and there’s a trillion dollars of digital capital in the banking system, it won’t just be in the U.S. It’s a virus. And so the virus spreads. And in this case, that means you’re going to have hundreds of thousands of banks and trillions of dollars that are held by a billion people.”

‘Thermodynamically Sound’ Money

Michael Saylor was born in Lincoln, Nebraska. He grew up on Air Force bases across the Midwest, as well as in Japan and New Zealand. An Air Force scholarship sent Saylor to the Massachusetts Institute of Technology, where he obtained dual degrees in aeronautics, astronautics, and the history of science. A literal rocket scientist, Saylor’s systems mindset attracted him to bitcoin’s “thermodynamically sound” design.

After serving as an Air Force Reserve captain, Saylor co-founded MicroStrategy in 1989, a software firm that rode the dot-com bubble, until Saylor and two other MicroStrategy executives were embroiled in an accounting fraud scandal in 2000. Eventually, they settled with the U.S. Securities and Exchange Commission for about $11 million.

At MicroStrategy, Saylor invented over 48 patents and deployed dozens of business ideas. Some succeeded, most of them failed. Saylor said the irony is that his greatest success was somebody else’s idea. Satoshi Namamoto, the pseudonymous creator of Bitcoin, created “digital gold” that Saylor discovered while under lockdown during the Covid-19 pandemic. He grabbed onto it out of desperation, preferring MicroStrategy to have a quick death over a slow one if it failed.

In July 2020, MicroStrategy began to steadily and continuously purchase bitcoin through cash flows, equity and debt, basically any way it could. It climbed the highs of the 2021 bull run and withstood the impairment charges of the 2022 crypto winter. By 2024, the Bitcoin corporate treasury strategy emerged battle tested. It survived its first crypto market cycle and the Trump bump catapulted MicroStrategy from a $1 billion to a $100 billion market cap company.

“[Bitcoin] became an opportunity,” said Saylor. “Then it became a strategy, and then all of a sudden in the past 12 months, we realized it was a really good business.”

From MicroStrategy to Strategy

MicroStrategy, rebranded and doing business as “Strategy,” proved to be an incredibly desirable stock for institutional investors wanting exposure to the volatile ups and downs of bitcoin. In December, Strategy was admitted to the Nasdaq 100. It is now eyeing membership to the S&P 500, which would spark an additional tidal wave of public market access.

To generate positive momentum, Strategy is laser-focused on raising capital to buy more bitcoin through a plethora of fixed income securities, creating a casino of financial products for traders addicted to bitcoin’s volatility. By constantly weighing market conditions, tweaking yield parameters and conversion factors, Strategy has engineered “intelligent leverage” designed to lure demand and ensure each successive series of securities amp each other up in an endless positive feedback loop.

“If you were to say, it sounds like financial engineering, it absolutely is financial engineering,” said Saylor. “ It creates more pressure to drive up the price of bitcoin, which drives up the price of MSTR, which drives up the leverage of MSTR, which drives up the value of the options, which drives up the demand for the equity, which drives up the demand and the value of the [convertible bonds], which drives up the price of and the demand for the preferred [shares].”

Strategy has raised approximately $33 billion to purchase half a billion bitcoins through this financial engineering. That has ignited online debate regarding Strategy’s ability to pay out dividends or bond maturities if markets sour or it cannot raise fresh capital. The money likely won’t come from existing company cash flows: Strategy’s software profits are negligible; in 2020-2023, they were negative, according to MarketWatch data.

All of this keeps Saylor up at night. So, Strategy is keeping all of its options open.

“ When the equity capital markets give us a massive premium, we’ll sell the equity,” said Saylor. “If we get too levered, we will de-lever. If we feel that the capital markets aren’t really favorable to sell any securities, we’ll just stop and wait.”

Last week, Strategy brought its bitcoin holdings above 500,000 tokens by purchasing an additional 6,911 bitcoins for $584 million, using proceeds from the sale of MSTR common stock. They further announced their new STRF perpetual offering raised $711 million to buy more bitcoin, when its initial goal was to raise $500 million.

This latest series of preferred stock differs from the original STRK offering in that it comes with a higher coupon (10% versus 8%) and has no common share conversion provision. Spelled out in the prospectuses of both offerings are risk factors that include no obligation to pay accumulated dividends “for any reason.”

Strategy has also eliminated any collateralized debt and therefore liquidation risk of the company’s bitcoin assets.

”We’ve built an indestructible balance sheet. Bitcoin could trade down 99%. There’s no margin call coming. The instruments that are constructed don’t have Bitcoin pledged as collateral,” said Saylor.

Ultimately, the dates to watch are when Strategy’s loans to bondholders become due. The first “put date” is September 16, 2027. If Strategy fails to incentivize bondholders to convert their bonds to MSTR stock or persuade them to await principal repayment the following year, these bondholders might demand Strategy buy back their $1.8 billion loan in cash. If the markets are still hungry for bitcoin exposure, it will be easier to raise capital and pay back investors. If there is a market downturn, and the Wall Street spigot runs dry, Strategy may have to consider selling its bitcoin or default.

‘Economic immortality’

But Saylor said Strategy, like the U.S. government, will “never sell” its bitcoin. He’s bet everything on BTC price going up forever, and the sovereignty, sound money, freedom, and property rights idealized by the community.

Before he dies, Saylor may burn bitcoin rather than give his assets away. That would be a “more ethically proper, ethically sound form of charity” and would bestow “economic immortality.”

“ If I believe that and I burn those keys, then I have made everybody in the [Bitcoin] network that much richer and more powerful forever,” said Saylor. “We’re all in it together, from now to eternity. So yeah, that’s my legacy.”

Elon Musk’s AI Start-Up and Nvidia Join Microsoft, BlackRock, MGX AI Fund

Two of the biggest forces in artificial intelligence (AI) — Elon Musk’s xAI and Nvidia — have joined BlackRock, Microsoft and investment fund MGX’s group to expand AI infrastructure across the U.S., the companies announced Wednesday.

The fund — called the AI Infrastructure Partnership — at its formation in September of last year, said that it plans to launch with more than $30 billion in initial funding. The goal is to build data centers and energy projects which are required to power large-scale AI models.

Nvidia will also serve as a technical advisor for the group, which it had announced last year.

Must and Nvidia join the effort two months after U.S. President Donald Trump announced the formation of Stargate, a private venture that plans to build up to 20 large AI data centers in the U.S. in partnership with OpenAI, Oracle and SoftBank.

The first data center will be built in Abilene, a small city in Texas, which will be completed by mid-2026, Bloomberg reported yesterday. It will have space for roughly 400,000 of Nvidia’s AI chips and a capacity of 1.2 gigawatts of power.

What’s the Impact of CoreWeave’s IPO on Core Scientific? Analysts Debate

CoreWeave, one of the hottest players in the artificial intelligence (AI) sphere, filed for an initial public offering (IPO) on March 3.

The company works closely with bitcoin (BTC) miner Core Scientific (CORZ); the two firms have inked a multi-billion dollar deal for Core Scientific to build hundreds of megawatts of infrastructure for CoreWeave to run its AI services. In fact, CoreWeave is by far Core Scientific’s largest client.

Which raises a question: How will CoreWeave’s IPO affect Core Scientific?

“CoreWeave’s IPO likely will have a correlating impact on Core Scientific,” Wolfie Zhao, head of research at TheMinerMag, told CoinDesk. “If CoreWeave’s public debut is successful and it sustains strong revenue growth in the coming years, it will reinforce Core Scientific’s position as a reliable infrastructure provider, ensuring a stable revenue stream from hosting CoreWeave’s GPUs.”

“However, if the AI market experiences a downturn or the demand for high-performance computing weakens, Core could face similar headwinds, as its business is increasingly tied to the broader AI ecosystem,” Zhao added.

The IPO may end up being less of a big deal than CoreWeave’s new $12 billion deal with AI heavyweight OpenAI, analysts at investment bank Canaccord Genuity suggested in a March 12 note. The deal will allow CoreWeave to diversify its revenue away from Microsoft, which currently contributes roughly two thirds of the firm’s revenue. That diversification should end up benefitting Core Scientific on account of the fact that CoreWeave is the firm’s biggest customer.

This quest for diversity is unlikely to hurt the relationship between CoreWeave and Core Scientific, analysts at investment firm H.C. Wainwright wrote on March 11. CoreWeave’s IPO is another sign that rumours that CoreWeave would seek to drop its contract with Core Scientific are untrue. “From the 20,000-foot level, how in the world would that make any sense for CoreWeave?” they wrote. Not only do the two firms have a long history of working together, but they regularly make agreements for Core Scientific to provide even more infrastructure to the AI Hyperscaler.

As for Core Scientific itself, the firm’s executive team is “thrilled” by the potential IPO, a spokesperson told CoinDesk. “We’re proud to be part of their journey toward becoming a public company and look forward to supporting their continued success. Nothing is more rewarding than seeing our customers grow, and we’re excited to continue scaling alongside them as they reach new milestones.

Blockchain Firm Crossmint Used by Adidas, Red Bull Raises $23.6M in Funding

Crossmint, a blockchain infrastructure firm helping companies build on-chain applications, has raised $23.6 million in funding.

The company, which has over 40,000 users, aims to simplify blockchain adoption by enabling firms to integrate wallets, tokenization, and payments with minimal code, according to a statement on Tuesday. Crossmint users, including big brands Adidas and Red Bull, use the platform to transition their operations on-chain.

Crossmint is also building a framework for artificial intelligence-driven commerce, providing wallets and payment APIs for AI agents.

“AI agents are reshaping commerce. Soon, they will autonomously manage tasks like grocery shopping or personal styling,” said Alfonso Gomez-Jordana, co-founder of Crossmint. “Traditional payment systems weren’t designed for AI agents—but blockchain is.”

Ribbit Capital led the investment round with additional participation from Franklin Templeton, Nyca, First Round, and Lightspeed Faction, Crossmint announced on Tuesday.

Disclaimer: This article, or parts of it, was generated with assistance from AI tools and reviewed by our editorial team to ensure accuracy and adherence to our standards. For more information, see CoinDesk’s full AI Policy.

Halliday Raises $20M for AI Protocol to Eliminate Writing Smart Contracts for DeFi

Artificial intelligence (AI)-focused blockchain protocol Halliday said it raised $20 million to help fund development of its Agentic Workflow Protocol (AWP), which aims to speed development of decentralized finance (DeFi) applications and avoid the need for programmers to write smart contracts.

The Series A funding round was led by venture capital giant Andreessen Horowitz’s (a16z) crypto arm.

“Our mission is to pioneer the software era of blockchain, enabling developers to build applications in hours, not years,” Halliday said in an emailed announcement. “With Halliday, you can never write a smart contract again.”

Halliday’s no-code blockchain automation could help to accelerate blockchain adoption and shift development from smart contracts to AI-driven workflows, making blockchain more accessible, scalable, and efficient.

The funding round, which follows Halliday’s seed $6 million round in 2022 — also led by a16z crypto — included contributions from Avalanche Blizzard Fund, Credibly Neutral and Alt Layer.

Disclaimer: This article, or parts of it, was generated with assistance from AI tools and reviewed by our editorial team to ensure accuracy and adherence to our standards. For more information, see CoinDesk’s full AI Policy.

Sam Altman’s World Network and Razer Want to Defeat Gaming’s Bot Problem

Sam Altman’s blockchain project, World Network, is teaming up with gaming hardware firm Razer on a suite of features designed to weed out bots from video games.

“Razer ID verified by World ID” is a single sign-on mechanism that will verify real human gamers from bots. It’s built atop Razer ID, Razer’s existing login service, and will help guarantee there’s “a real person behind every Razer ID account,” according to a statement shared by Razer and World.

The collaboration between the two firms comes as artificial intelligence (AI) tools are seeping into every corner of online life — including inside of video games, which have been plagued by non-human AI “bots” since long before the rise of Altman’s ChatGPT.

According to a study from Echelon Insights that World shared with CoinDesk, roughly 59% of gamers said that they regularly encountered unauthorized, third-party bots in their games. In addition to posing a general nuisance to players, bot accounts often have tactical advantages over real players, which can ruin the competitiveness of some multiplayer games.

“Game developers now have a tool to build dynamic spaces where real players —not bots— dominate the digital landscape,” World said in its statement.

Razer’s integration with World Network builds upon World’s existing blockchain-based identity solution, which uses iris scans to differentiate real humans from robots online.

The new feature will be integrated first into “TOKYO BEAST,” a blockchain-based game set in a version of Tokyo based 100 years in the future. It’s an apt pairing: the game’s main premise involves humans coexisting with autonomous androids.

When users log into TOKYO BEAST, they will be prompted to sign in using a World-authenticated Razer ID, ensuring they can play online with real human players only.

“As AI continues to reshape the gaming world, we want to empower gamers and game developers with the tools they need to navigate this transformation safely and confidently,” said Wei-Pin Choo, the chief corporate officer at Razer. “By teaming up with World, we’re ensuring that real players are the heart of every experience, keeping gaming fair, immersive, and designed for humans.”

Read more: Sam Altman’s World Network Unveils New Chat Feature to Connect Real Humans

Can AI bots steal your crypto? The rise of digital thieves

Can AI bots steal your crypto? The rise of digital thieves

What are AI bots?

AI bots are self-learning software that automates and continuously refines crypto cyberattacks, making them more dangerous than traditional hacking methods.

At the heart of today’s AI-driven cybercrime are AI bots — self-learning software programs designed to process vast amounts of data, make independent decisions, and execute complex tasks without human intervention. While these bots have been a game-changer in industries like finance, healthcare and customer service, they have also become a weapon for cybercriminals, particularly in the world of cryptocurrency.

Unlike traditional hacking methods, which require manual effort and technical expertise, AI bots can fully automate attacks, adapt to new cryptocurrency security measures, and even refine their tactics over time. This makes them far more effective than human hackers, who are limited by time, resources and error-prone processes.

Why are AI bots so dangerous?

The biggest threat posed by AI-driven cybercrime is scale. A single hacker attempting to breach a crypto exchange or trick users into handing over their private keys can only do so much. AI bots, however, can launch thousands of attacks simultaneously, refining their techniques as they go.

  • Speed: AI bots can scan millions of blockchain transactions, smart contracts and websites within minutes, identifying weaknesses in wallets (leading to crypto wallet hacks), decentralized finance (DeFi) protocols and exchanges.
  • Scalability: A human scammer may send phishing emails to a few hundred people. An AI bot can send personalized, perfectly crafted phishing emails to millions in the same time frame.
  • Adaptability: Machine learning allows these bots to improve with every failed attack, making them harder to detect and block.

This ability to automate, adapt and attack at scale has led to a surge in AI-driven crypto fraud, making crypto fraud prevention more critical than ever.

​In October 2024, the X account of Andy Ayrey, developer of the AI bot Truth Terminal, was compromised by hackers. The attackers used Ayrey’s account to promote a fraudulent memecoin named Infinite Backrooms (IB). The malicious campaign led to a rapid surge in IB’s market capitalization, reaching $25 million. Within 45 minutes, the perpetrators liquidated their holdings, securing over $600,000.

How AI-powered bots can steal cryptocurrency assets

AI-powered bots aren’t just automating crypto scams — they’re becoming smarter, more targeted and increasingly hard to spot.

Here are some of the most dangerous types of AI-driven scams currently being used to steal cryptocurrency assets:

1. AI-powered phishing bots

Phishing attacks are nothing new in crypto, but AI has turned them into a far bigger threat. Instead of sloppy emails full of mistakes, today’s AI bots create personalized messages that look exactly like real communications from platforms such as Coinbase or MetaMask. They gather personal information from leaked databases, social media and even blockchain records, making their scams extremely convincing. 

For instance, in early 2024, an AI-driven phishing attack targeted Coinbase users by sending emails about fake cryptocurrency security alerts, ultimately tricking users out of nearly $65 million.

Also, after OpenAI launched GPT-4, scammers created a fake OpenAI token airdrop site to exploit the hype. They sent emails and X posts luring users to “claim” a bogus token — the phishing page closely mirrored OpenAI’s real site​. Victims who took the bait and connected their wallets had all their crypto assets drained automatically.

Unlike old-school phishing, these AI-enhanced scams are polished and targeted, often free of the typos or clumsy wording that is used to give away a phishing scam. Some even deploy AI chatbots posing as customer support representatives for exchanges or wallets, tricking users into divulging private keys or two-factor authentication (2FA) codes under the guise of “verification.”

In 2022, some malware specifically targeted browser-based wallets like MetaMask: a strain called Mars Stealer could sniff out private keys for over 40 different wallet browser extensions and 2FA apps, draining any funds it found. Such malware often spreads via phishing links, fake software downloads or pirated crypto tools.

Once inside your system, it might monitor your clipboard (to swap in the attacker’s address when you copy-paste a wallet address), log your keystrokes, or export your seed phrase files — all without obvious signs.

2. AI-powered exploit-scanning bots

Smart contract vulnerabilities are a hacker’s goldmine, and AI bots are taking advantage faster than ever. These bots continuously scan platforms like Ethereum or BNB Smart Chain, hunting for flaws in newly deployed DeFi projects. As soon as they detect an issue, they exploit it automatically, often within minutes. 

Researchers have demonstrated that AI chatbots, such as those powered by GPT-3, can analyze smart contract code to identify exploitable weaknesses. For instance, Stephen Tong, co-founder of Zellic, showcased an AI chatbot detecting a vulnerability in a smart contract’s “withdraw” function, similar to the flaw exploited in the Fei Protocol attack, which resulted in an $80-million loss. 

3. AI-enhanced brute-force attacks

Brute-force attacks used to take forever, but AI bots have made them dangerously efficient. By analyzing previous password breaches, these bots quickly identify patterns to crack passwords and seed phrases in record time. A 2024 study on desktop cryptocurrency wallets, including Sparrow, Etherwall and Bither, found that weak passwords drastically lower resistance to brute-force attacks, emphasizing that strong, complex passwords are crucial to safeguarding digital assets.

4. Deepfake impersonation bots

Imagine watching a video of a trusted crypto influencer or CEO asking you to invest — but it’s entirely fake. That’s the reality of deepfake scams powered by AI. These bots create ultra-realistic videos and voice recordings, tricking even savvy crypto holders into transferring funds. 

Horrifying ripple scam on social media

5. Social media botnets

On platforms like X and Telegram, swarms of AI bots push crypto scams at scale. Botnets such as “Fox8” used ChatGPT to generate hundreds of persuasive posts hyping scam tokens and replying to users in real-time.

In one case, scammers abused the names of Elon Musk and ChatGPT to promote a fake crypto giveaway — complete with a deepfaked video of Musk — duping people into sending funds to scammers. 

In 2023, Sophos researchers found crypto romance scammers using ChatGPT to chat with multiple victims at once, making their affectionate messages more convincing and scalable.​

How the scammer used large language model-based AI in chat responses

Similarly, Meta reported a sharp uptick in malware and phishing links disguised as ChatGPT or AI tools, often tied to crypto fraud schemes. And in the realm of romance scams, AI is boosting so-called pig butchering operations — long-con scams where fraudsters cultivate relationships and then lure victims into fake crypto investments. A striking case occurred in Hong Kong in 2024: Police busted a criminal ring that defrauded men across Asia of $46 million via an AI-assisted romance scam​.

Automated trading bot scams and exploits

AI is being invoked in the arena of cryptocurrency trading bots — often as a buzzword to con investors and occasionally as a tool for technical exploits.

A notable example is YieldTrust.ai, which in 2023 marketed an AI bot supposedly yielding 2.2% returns per day — an astronomical, implausible profit. Regulators from several states investigated and found no evidence the “AI bot” even existed; it appeared to be a classic Ponzi, using AI as a tech buzzword to suck in victims​. YieldTrust.ai was ultimately shut down by authorities, but not before investors were duped by the slick marketing. 

Even when an automated trading bot is real, it’s often not the money-printing machine scammers claim. For instance, blockchain analysis firm Arkham Intelligence highlighted a case where a so-called arbitrage trading bot (likely touted as AI-driven) executed an incredibly complex series of trades, including a $200-million flash loan — and ended up netting a measly $3.24 in profit​.

In fact, many “AI trading” scams will take your deposit and, at best, run it through some random trades (or not trade at all), then make excuses when you try to withdraw. Some shady operators also use social media AI bots to fabricate a track record (e.g., fake testimonials or X bots that constantly post “winning trades”) to create an illusion of success. It’s all part of the ruse.

On the more technical side, criminals do use automated bots (not necessarily AI, but sometimes labeled as such) to exploit the crypto markets and infrastructure. Front-running bots in DeFi, for example, automatically insert themselves into pending transactions to steal a bit of value (a sandwich attack), and flash loan bots execute lightning-fast trades to exploit price discrepancies or vulnerable smart contracts. These require coding skills and aren’t typically marketed to victims; instead, they’re direct theft tools used by hackers. 

AI could enhance these by optimizing strategies faster than a human. However, as mentioned, even highly sophisticated bots don’t guarantee big gains — the markets are competitive and unpredictable, something even the fanciest AI can’t reliably foresee​.

Meanwhile, the risk to victims is real: If a trading algorithm malfunctions or is maliciously coded, it can wipe out your funds in seconds. There have been cases of rogue bots on exchanges triggering flash crashes or draining liquidity pools, causing users to incur huge slippage losses.

How AI-powered malware fuels cybercrime against crypto users

AI is teaching cybercriminals how to hack crypto platforms, enabling a wave of less-skilled attackers to launch credible attacks. This helps explain why crypto phishing and malware campaigns have scaled up so dramatically — AI tools let bad actors automate their scams and continuously refine them based on what works​.

AI is also supercharging malware threats and hacking tactics aimed at crypto users. One concern is AI-generated malware, malicious programs that use AI to adapt and evade detection. 

In 2023, researchers demonstrated a proof-of-concept called BlackMamba, a polymorphic keylogger that uses an AI language model (like the tech behind ChatGPT) to rewrite its code with every execution. This means each time BlackMamba runs, it produces a new variant of itself in memory, helping it slip past antivirus and endpoint security tools​.

​In tests, this AI-crafted malware went undetected by an industry-leading endpoint detection and response system​. Once active, it could stealthily capture everything the user types — including crypto exchange passwords or wallet seed phrases — and send that data to attackers​.

While BlackMamba was just a lab demo, it highlights a real threat: Criminals can harness AI to create shape-shifting malware that targets cryptocurrency accounts and is much harder to catch than traditional viruses​.

Even without exotic AI malware, threat actors abuse the popularity of AI to spread classic trojans. Scammers commonly set up fake “ChatGPT” or AI-related apps that contain malware, knowing users might drop their guard due to the AI branding. For instance, security analysts observed fraudulent websites impersonating the ChatGPT site with a “Download for Windows” button; if clicked, it silently installs a crypto-stealing Trojan on the victim’s machine​.

Beyond the malware itself, AI is lowering the skill barrier for would-be hackers. Previously, a criminal needed some coding know-how to craft phishing pages or viruses. Now, underground “AI-as-a-service” tools do much of the work. 

Illicit AI chatbots like WormGPT and FraudGPT have appeared on dark web forums, offering to generate phishing emails, malware code and hacking tips on demand​. For a fee, even non-technical criminals can use these AI bots to churn out convincing scam sites, create new malware variants, and scan for software vulnerabilities​.

How to protect your crypto from AI-driven attacks

AI-driven threats are becoming more advanced, making strong security measures essential to protect digital assets from automated scams and hacks.

Below are the most effective ways on how to protect crypto from hackers and defend against AI-powered phishing, deepfake scams and exploit bots:

  • Use a hardware wallet: AI-driven malware and phishing attacks primarily target online (hot) wallets. By using hardware wallets — like Ledger or Trezor — you keep private keys completely offline, making them virtually impossible for hackers or malicious AI bots to access remotely. For instance, during the 2022 FTX collapse, those using hardware wallets avoided the massive losses suffered by users with funds stored on exchanges.
  • Enable multifactor authentication (MFA) and strong passwords: AI bots can crack weak passwords using deep learning in cybercrime, leveraging machine learning algorithms trained on leaked data breaches to predict and exploit vulnerable credentials. To counter this, always enable MFA via authenticator apps like Google Authenticator or Authy rather than SMS-based codes — hackers have been known to exploit SIM swap vulnerabilities, making SMS verification less secure.
  • Beware of AI-powered phishing scams: AI-generated phishing emails, messages and fake support requests have become nearly indistinguishable from real ones. Avoid clicking on links in emails or direct messages, always verify website URLs manually, and never share private keys or seed phrases, regardless of how convincing the request may seem.
  • Verify identities carefully to avoid deepfake scams: AI-powered deepfake videos and voice recordings can convincingly impersonate crypto influencers, executives or even people you personally know. If someone is asking for funds or promoting an urgent investment opportunity via video or audio, verify their identity through multiple channels before taking action.
  • Stay informed about the latest blockchain security threats: Regularly following trusted blockchain security sources such as CertiK, Chainalysis or SlowMist will keep you informed about the latest AI-powered threats and the tools available to protect yourself. 

The future of AI in cybercrime and crypto security

As AI-driven crypto threats evolve rapidly, proactive and AI-powered security solutions become crucial to protecting your digital assets.

Looking ahead, AI’s role in cybercrime is likely to escalate, becoming increasingly sophisticated and harder to detect. Advanced AI systems will automate complex cyberattacks like deepfake-based impersonations, exploit smart-contract vulnerabilities instantly upon detection, and execute precision-targeted phishing scams. 

To counter these evolving threats, blockchain security will increasingly rely on real-time AI threat detection. Platforms like CertiK already leverage advanced machine learning models to scan millions of blockchain transactions daily, spotting anomalies instantly. 

As cyber threats grow smarter, these proactive AI systems will become essential in preventing major breaches, reducing financial losses, and combating AI and financial fraud to maintain trust in crypto markets.

Ultimately, the future of crypto security will depend heavily on industry-wide cooperation and shared AI-driven defense systems. Exchanges, blockchain platforms, cybersecurity providers and regulators must collaborate closely, using AI to predict threats before they materialize. While AI-powered cyberattacks will continue to evolve, the crypto community’s best defense is staying informed, proactive and adaptive — turning artificial intelligence from a threat into its strongest ally.

Can AI bots steal your crypto? The rise of digital thieves

Can AI bots steal your crypto? The rise of digital thieves

What are AI bots?

AI bots are self-learning software that automates and continuously refines crypto cyberattacks, making them more dangerous than traditional hacking methods.

At the heart of today’s AI-driven cybercrime are AI bots — self-learning software programs designed to process vast amounts of data, make independent decisions, and execute complex tasks without human intervention. While these bots have been a game-changer in industries like finance, healthcare and customer service, they have also become a weapon for cybercriminals, particularly in the world of cryptocurrency.

Unlike traditional hacking methods, which require manual effort and technical expertise, AI bots can fully automate attacks, adapt to new cryptocurrency security measures, and even refine their tactics over time. This makes them far more effective than human hackers, who are limited by time, resources and error-prone processes.

Why are AI bots so dangerous?

The biggest threat posed by AI-driven cybercrime is scale. A single hacker attempting to breach a crypto exchange or trick users into handing over their private keys can only do so much. AI bots, however, can launch thousands of attacks simultaneously, refining their techniques as they go.

  • Speed: AI bots can scan millions of blockchain transactions, smart contracts and websites within minutes, identifying weaknesses in wallets (leading to crypto wallet hacks), decentralized finance (DeFi) protocols and exchanges.
  • Scalability: A human scammer may send phishing emails to a few hundred people. An AI bot can send personalized, perfectly crafted phishing emails to millions in the same time frame.
  • Adaptability: Machine learning allows these bots to improve with every failed attack, making them harder to detect and block.

This ability to automate, adapt and attack at scale has led to a surge in AI-driven crypto fraud, making crypto fraud prevention more critical than ever.

​In October 2024, the X account of Andy Ayrey, developer of the AI bot Truth Terminal, was compromised by hackers. The attackers used Ayrey’s account to promote a fraudulent memecoin named Infinite Backrooms (IB). The malicious campaign led to a rapid surge in IB’s market capitalization, reaching $25 million. Within 45 minutes, the perpetrators liquidated their holdings, securing over $600,000.

How AI-powered bots can steal cryptocurrency assets

AI-powered bots aren’t just automating crypto scams — they’re becoming smarter, more targeted and increasingly hard to spot.

Here are some of the most dangerous types of AI-driven scams currently being used to steal cryptocurrency assets:

1. AI-powered phishing bots

Phishing attacks are nothing new in crypto, but AI has turned them into a far bigger threat. Instead of sloppy emails full of mistakes, today’s AI bots create personalized messages that look exactly like real communications from platforms such as Coinbase or MetaMask. They gather personal information from leaked databases, social media and even blockchain records, making their scams extremely convincing. 

For instance, in early 2024, an AI-driven phishing attack targeted Coinbase users by sending emails about fake cryptocurrency security alerts, ultimately tricking users out of nearly $65 million.

Also, after OpenAI launched GPT-4, scammers created a fake OpenAI token airdrop site to exploit the hype. They sent emails and X posts luring users to “claim” a bogus token — the phishing page closely mirrored OpenAI’s real site​. Victims who took the bait and connected their wallets had all their crypto assets drained automatically.

Unlike old-school phishing, these AI-enhanced scams are polished and targeted, often free of the typos or clumsy wording that is used to give away a phishing scam. Some even deploy AI chatbots posing as customer support representatives for exchanges or wallets, tricking users into divulging private keys or two-factor authentication (2FA) codes under the guise of “verification.”

In 2022, some malware specifically targeted browser-based wallets like MetaMask: a strain called Mars Stealer could sniff out private keys for over 40 different wallet browser extensions and 2FA apps, draining any funds it found. Such malware often spreads via phishing links, fake software downloads or pirated crypto tools.

Once inside your system, it might monitor your clipboard (to swap in the attacker’s address when you copy-paste a wallet address), log your keystrokes, or export your seed phrase files — all without obvious signs.

2. AI-powered exploit-scanning bots

Smart contract vulnerabilities are a hacker’s goldmine, and AI bots are taking advantage faster than ever. These bots continuously scan platforms like Ethereum or BNB Smart Chain, hunting for flaws in newly deployed DeFi projects. As soon as they detect an issue, they exploit it automatically, often within minutes. 

Researchers have demonstrated that AI chatbots, such as those powered by GPT-3, can analyze smart contract code to identify exploitable weaknesses. For instance, Stephen Tong, co-founder of Zellic, showcased an AI chatbot detecting a vulnerability in a smart contract’s “withdraw” function, similar to the flaw exploited in the Fei Protocol attack, which resulted in an $80-million loss. 

3. AI-enhanced brute-force attacks

Brute-force attacks used to take forever, but AI bots have made them dangerously efficient. By analyzing previous password breaches, these bots quickly identify patterns to crack passwords and seed phrases in record time. A 2024 study on desktop cryptocurrency wallets, including Sparrow, Etherwall and Bither, found that weak passwords drastically lower resistance to brute-force attacks, emphasizing that strong, complex passwords are crucial to safeguarding digital assets.

4. Deepfake impersonation bots

Imagine watching a video of a trusted crypto influencer or CEO asking you to invest — but it’s entirely fake. That’s the reality of deepfake scams powered by AI. These bots create ultra-realistic videos and voice recordings, tricking even savvy crypto holders into transferring funds. 

Horrifying ripple scam on social media

5. Social media botnets

On platforms like X and Telegram, swarms of AI bots push crypto scams at scale. Botnets such as “Fox8” used ChatGPT to generate hundreds of persuasive posts hyping scam tokens and replying to users in real-time.

In one case, scammers abused the names of Elon Musk and ChatGPT to promote a fake crypto giveaway — complete with a deepfaked video of Musk — duping people into sending funds to scammers. 

In 2023, Sophos researchers found crypto romance scammers using ChatGPT to chat with multiple victims at once, making their affectionate messages more convincing and scalable.​

How the scammer used large language model-based AI in chat responses

Similarly, Meta reported a sharp uptick in malware and phishing links disguised as ChatGPT or AI tools, often tied to crypto fraud schemes. And in the realm of romance scams, AI is boosting so-called pig butchering operations — long-con scams where fraudsters cultivate relationships and then lure victims into fake crypto investments. A striking case occurred in Hong Kong in 2024: Police busted a criminal ring that defrauded men across Asia of $46 million via an AI-assisted romance scam​.

Automated trading bot scams and exploits

AI is being invoked in the arena of cryptocurrency trading bots — often as a buzzword to con investors and occasionally as a tool for technical exploits.

A notable example is YieldTrust.ai, which in 2023 marketed an AI bot supposedly yielding 2.2% returns per day — an astronomical, implausible profit. Regulators from several states investigated and found no evidence the “AI bot” even existed; it appeared to be a classic Ponzi, using AI as a tech buzzword to suck in victims​. YieldTrust.ai was ultimately shut down by authorities, but not before investors were duped by the slick marketing. 

Even when an automated trading bot is real, it’s often not the money-printing machine scammers claim. For instance, blockchain analysis firm Arkham Intelligence highlighted a case where a so-called arbitrage trading bot (likely touted as AI-driven) executed an incredibly complex series of trades, including a $200-million flash loan — and ended up netting a measly $3.24 in profit​.

In fact, many “AI trading” scams will take your deposit and, at best, run it through some random trades (or not trade at all), then make excuses when you try to withdraw. Some shady operators also use social media AI bots to fabricate a track record (e.g., fake testimonials or X bots that constantly post “winning trades”) to create an illusion of success. It’s all part of the ruse.

On the more technical side, criminals do use automated bots (not necessarily AI, but sometimes labeled as such) to exploit the crypto markets and infrastructure. Front-running bots in DeFi, for example, automatically insert themselves into pending transactions to steal a bit of value (a sandwich attack), and flash loan bots execute lightning-fast trades to exploit price discrepancies or vulnerable smart contracts. These require coding skills and aren’t typically marketed to victims; instead, they’re direct theft tools used by hackers. 

AI could enhance these by optimizing strategies faster than a human. However, as mentioned, even highly sophisticated bots don’t guarantee big gains — the markets are competitive and unpredictable, something even the fanciest AI can’t reliably foresee​.

Meanwhile, the risk to victims is real: If a trading algorithm malfunctions or is maliciously coded, it can wipe out your funds in seconds. There have been cases of rogue bots on exchanges triggering flash crashes or draining liquidity pools, causing users to incur huge slippage losses.

How AI-powered malware fuels cybercrime against crypto users

AI is teaching cybercriminals how to hack crypto platforms, enabling a wave of less-skilled attackers to launch credible attacks. This helps explain why crypto phishing and malware campaigns have scaled up so dramatically — AI tools let bad actors automate their scams and continuously refine them based on what works​.

AI is also supercharging malware threats and hacking tactics aimed at crypto users. One concern is AI-generated malware, malicious programs that use AI to adapt and evade detection. 

In 2023, researchers demonstrated a proof-of-concept called BlackMamba, a polymorphic keylogger that uses an AI language model (like the tech behind ChatGPT) to rewrite its code with every execution. This means each time BlackMamba runs, it produces a new variant of itself in memory, helping it slip past antivirus and endpoint security tools​.

​In tests, this AI-crafted malware went undetected by an industry-leading endpoint detection and response system​. Once active, it could stealthily capture everything the user types — including crypto exchange passwords or wallet seed phrases — and send that data to attackers​.

While BlackMamba was just a lab demo, it highlights a real threat: Criminals can harness AI to create shape-shifting malware that targets cryptocurrency accounts and is much harder to catch than traditional viruses​.

Even without exotic AI malware, threat actors abuse the popularity of AI to spread classic trojans. Scammers commonly set up fake “ChatGPT” or AI-related apps that contain malware, knowing users might drop their guard due to the AI branding. For instance, security analysts observed fraudulent websites impersonating the ChatGPT site with a “Download for Windows” button; if clicked, it silently installs a crypto-stealing Trojan on the victim’s machine​.

Beyond the malware itself, AI is lowering the skill barrier for would-be hackers. Previously, a criminal needed some coding know-how to craft phishing pages or viruses. Now, underground “AI-as-a-service” tools do much of the work. 

Illicit AI chatbots like WormGPT and FraudGPT have appeared on dark web forums, offering to generate phishing emails, malware code and hacking tips on demand​. For a fee, even non-technical criminals can use these AI bots to churn out convincing scam sites, create new malware variants, and scan for software vulnerabilities​.

How to protect your crypto from AI-driven attacks

AI-driven threats are becoming more advanced, making strong security measures essential to protect digital assets from automated scams and hacks.

Below are the most effective ways on how to protect crypto from hackers and defend against AI-powered phishing, deepfake scams and exploit bots:

  • Use a hardware wallet: AI-driven malware and phishing attacks primarily target online (hot) wallets. By using hardware wallets — like Ledger or Trezor — you keep private keys completely offline, making them virtually impossible for hackers or malicious AI bots to access remotely. For instance, during the 2022 FTX collapse, those using hardware wallets avoided the massive losses suffered by users with funds stored on exchanges.
  • Enable multifactor authentication (MFA) and strong passwords: AI bots can crack weak passwords using deep learning in cybercrime, leveraging machine learning algorithms trained on leaked data breaches to predict and exploit vulnerable credentials. To counter this, always enable MFA via authenticator apps like Google Authenticator or Authy rather than SMS-based codes — hackers have been known to exploit SIM swap vulnerabilities, making SMS verification less secure.
  • Beware of AI-powered phishing scams: AI-generated phishing emails, messages and fake support requests have become nearly indistinguishable from real ones. Avoid clicking on links in emails or direct messages, always verify website URLs manually, and never share private keys or seed phrases, regardless of how convincing the request may seem.
  • Verify identities carefully to avoid deepfake scams: AI-powered deepfake videos and voice recordings can convincingly impersonate crypto influencers, executives or even people you personally know. If someone is asking for funds or promoting an urgent investment opportunity via video or audio, verify their identity through multiple channels before taking action.
  • Stay informed about the latest blockchain security threats: Regularly following trusted blockchain security sources such as CertiK, Chainalysis or SlowMist will keep you informed about the latest AI-powered threats and the tools available to protect yourself. 

The future of AI in cybercrime and crypto security

As AI-driven crypto threats evolve rapidly, proactive and AI-powered security solutions become crucial to protecting your digital assets.

Looking ahead, AI’s role in cybercrime is likely to escalate, becoming increasingly sophisticated and harder to detect. Advanced AI systems will automate complex cyberattacks like deepfake-based impersonations, exploit smart-contract vulnerabilities instantly upon detection, and execute precision-targeted phishing scams. 

To counter these evolving threats, blockchain security will increasingly rely on real-time AI threat detection. Platforms like CertiK already leverage advanced machine learning models to scan millions of blockchain transactions daily, spotting anomalies instantly. 

As cyber threats grow smarter, these proactive AI systems will become essential in preventing major breaches, reducing financial losses, and combating AI and financial fraud to maintain trust in crypto markets.

Ultimately, the future of crypto security will depend heavily on industry-wide cooperation and shared AI-driven defense systems. Exchanges, blockchain platforms, cybersecurity providers and regulators must collaborate closely, using AI to predict threats before they materialize. While AI-powered cyberattacks will continue to evolve, the crypto community’s best defense is staying informed, proactive and adaptive — turning artificial intelligence from a threat into its strongest ally.

AI’s Lead Over Crypto for VC Dollars Increased in Q1’25, But Does This Race Really Matter?

Crypto venture funding in the U.S. clocked in at approximately $861 million for the first three months of 2025, but was dwarfed by artificial intelligence’s nearly $20 billion haul, according to data provided by Pitchbook, showing how investors continue to pivot towards AI.

Data shows that investors closed 795 deals in the U.S in AI from January to March, with blockbuster deals like Databricks’ $15.3 billion round and Anthropic’s $2 billion raise dominating headlines.

Crypto’s largest blockbuster deal, in comparison, was Abu Dhabi’s MGX, with a $2 billion investment into Binance – the first institutional placement in the crypto exchange. Other deals of note include a $82 million raise from payment infrastructure company Mesh, ETF issuer Bitwise’s $70 million round, and digital asset bank Sygnum’s $58 million offering.

Prior reporting by Pitchbook shows that AI startups attracted one-third of global VC investment in 2024, totaling $131.5 billion, with nearly a quarter of new startups being an AI company across 4,318 VC deals, compared to crypto’s $4.9 billion across just 706 deals.

Analysis: Has AI stolen crypto’s venture dollars?

Blockbuster rounds from VCs in the AI space and headline-grabbing antics, such as OpenAI’s Sam Altman seeking trillions, and AI’s rise from technological novelty to household name thanks to transformer models, would make one think that there’s suddenly an investor preference for one over the other.

Historically, all data shows that VCs have generally favored AI over crypto, with AI and machine learning attracting consistent funding that’s expanded exponentially, according to Statista data, growing from $670 million in 2011 to $36 billion in 2020.

There’s only been one year where crypto beat AI for funding, and that comes with a caveat: narrower AI categorizations, like ABI Research‘s $22.3 billion AI estimate in 2021, suggest crypto briefly outpaced AI funding during the bullish crypto cycle before AI funding surged again to over $100 billion by 2024.

Keep in mind that all of this ignores crypto-native quirks like airdrops, which put fresh capital in the hands of users and, in turn, pump the token price, inflating the size of projects’ treasuries.

A recent report from Dragonfly found that between 2020 and 2024, the 11 largest airdrops generated $7 billion. This won’t close the gap between AI and crypto, but it shows that there are more ways to get a dollar than traditional venture capital.