Crypto security teams are facing a new imbalance: while AI model developers are restricting their most capable cyber-related systems, only a small number of crypto firms appear to have gained early access to those “frontier” tools.

Coinbase has said it secured access to Anthropic’s restricted Mythos model, and Zcash co-founder Zooko Wilcox has described how Anthropic used Mythos to audit the Zcash protocol at the request of Shielded Labs. Meanwhile, Binance’s chief security officer Jimmy Su told Cointelegraph that the exchange has been trying to make progress but has not obtained the most advanced frontier model.

Key takeaways

• Only select crypto companies have reportedly received early access to restricted frontier AI models used for cybersecurity work.
• Executives argue that gating advanced models may be necessary at first, but maintaining restrictions could become harder to justify as capabilities converge with public releases.
• Uneven access may widen the security gap between defenders and attackers, particularly as AI-assisted exploit workflows reportedly speed up.
• Some crypto-adjacent organizations, such as those embedded in critical infrastructure or security tooling, have also joined gated programs.

Why restricted “frontier” models are hard to distribute

The core issue is not whether AI can help security—many teams already use mainstream models for testing and review—but whether defenders get access to the most cyber-capable systems under developer guardrails.

Anthropic has stated that Mythos 5 shares the same underlying model as its publicly available Fable 5, but operates without safeguards that limit sensitive cybersecurity use. OpenAI is described as running a similar tiered approach, with a “Trusted Access for Cyber” pathway for verified defenders and a more permissive cyber-oriented version reserved for a smaller group conducting authorized penetration testing.

Crypto security executives interviewed by Cointelegraph suggested this kind of restricted rollout is likely warranted initially. However, they also highlighted a growing tension: once publicly available models begin to approach the same practical capabilities, continuous gating may become harder to defend—especially if attackers can leverage comparable tools from elsewhere.

Crypto executives push for faster verification pathways

Jimmy Su said Anthropic’s controlled release can be responsible because attackers may benefit from newly released capabilities sooner than defenders. In his framing, limiting early access can reduce the “blast radius,” at least during an initial testing period.

Solana Foundation’s chief information security officer Michael Coates supported guardrails but argued that “legitimate defenders” need a faster route to the models. He said the process should streamline verification and acceptance programs so security teams can use the best available systems to match the pace of exploitation.

Blockchain Capital’s Sean Cheetham expressed a similar long-term view. While restrictions can help avoid immediate misuse, broader availability could ultimately benefit defense because the population of legitimate security researchers is typically far larger than the small groups able to run highly sophisticated attacks. That scale dynamic—more defenders than adversaries—may flip the risk calculus over time.

Who has access—and who appears to be waiting

Despite being the world’s largest exchange by daily trading volume, Binance has not reportedly secured access to Mythos, according to Su. The exchange’s scale underscores the potential operational impact: Binance holds substantial assets on its platform, and a lack of frontier defensive tooling could leave major ecosystems to rely on less capable alternatives.

Beyond exchanges, other organizations have taken different approaches. Cointelegraph previously reported that Fireblocks, which provides custody and security services at large scale, sought access to Mythos but at the time relied on Anthropic’s publicly available model for pentesting. Cointelegraph also cited that Uniswap founder Hayden Adams criticized the safeguards on Fable 5 around cybersecurity prompts earlier this year.

The Ethereum Foundation has said it has been using “coordinated AI agents” to identify bugs across its systems, without disclosing which models were used. Cointelegraph reached out to the Ethereum Foundation, Fireblocks, and Uniswap to confirm whether they had received access to frontier restricted models since then.

Some crypto-adjacent companies, however, have moved further into gated programs. FIS—an infrastructure provider that partnered with Circle for USDC payments functionality last year—reportedly joined Anthropic’s Project Glasswing last month. Project Glasswing is described as Anthropic’s gated program for vetted cyber defenders and organizations responsible for critical software infrastructure to access restricted Mythos models.

HackerOne, which supports bug bounty and security testing for major organizations including crypto exchanges, also said it joined Project Glasswing, though its testing is confined to its own infrastructure rather than customer programs. Separately, Cointelegraph reached out to OpenAI and Anthropic to ask how many crypto firms had received access to restricted models.

AI-assisted attacks are reported to be accelerating

The access gap matters because defenders are not operating in a static threat environment. Cointelegraph reported that Boltz, a Bitcoin swap service, chose to halt its non-custodial bridge after observing a steady rise in AI-assisted hacking attempts over the prior few months. Boltz said attackers are now iterating faster than a team its size can find and patch, pointing to a practical pressure on incident response and code review cycles.

Other security events also suggest attackers are applying automation to find real weaknesses. Cointelegraph reported that Coinkite disclosed a vulnerability affecting some Coldcard devices, where a flaw in wallet seed generation produced less randomness than expected. Coinkite speculated that the attacker could have used AI to review earlier firmware versions to locate and exploit the weakness—even though the company had used what it described as one of the best available AI models to review its code in the weeks before.

Taken together, these reports support a broader concern: even strong internal testing using public AI tools may not be enough if adversaries deploy higher-end capabilities and iterate faster than teams can remediate.

What to watch next for crypto security

The immediate question is whether restricted model access will widen beyond early adopters and whether developers can design guardrails that protect the ecosystem without bottlenecking legitimate defenders. As AI capabilities diffuse—through both public releases and competing models—crypto teams will likely watch not only for new access announcements, but also for changes in exploitation tempo and patch turnaround times across major platforms.

Jump Capital has closed its eighth institutional fund with $350 million in capital commitments, with the new vehicle dedicated to backing founders building AI applications, enterprise infrastructure, and cybersecurity technologies.

In an official announcement, the venture firm said its latest investment strategy is built around three developments shaping enterprise AI adoption: software applications being rebuilt around AI, new infrastructure needed to support production-scale deployment, and cybersecurity adapting to increasingly autonomous systems. 

The firm said the new fund will continue its long-standing approach of investing in technical founders solving complex infrastructure and enterprise challenges.

Jump Capital sees three areas driving enterprise AI

According to the announcement, enterprise interest in artificial intelligence has moved beyond experimentation, with organizations now facing challenges around deploying AI reliably in production. Jump Capital said enterprises are struggling to realize meaningful returns because infrastructure and security have not advanced as quickly as AI adoption.

The firm said it expects one of the biggest opportunities over the coming years to come from rebuilding the enterprise technology stack to support AI-native software.

On the application side, Jump Capital believes AI is allowing software to execute work that previously depended on consultants, analysts, and other specialists. It said the next generation of enterprise software will become embedded into critical workflows while accumulating proprietary context that improves decision-making over time.

Beneath those applications, the firm said enterprise infrastructure remains in the early stages of development despite significant investment in data centers and graphics processors. It is evaluating startups building autonomous data engineering platforms, semantic and context layers, agent observability and governance, AI-native software quality, distributed inference systems, and orchestration platforms capable of managing increasingly complex AI environments.

Security forms the third pillar of the firm’s investment thesis. Jump Capital said AI adoption is expanding enterprise attack surfaces while introducing new challenges around agent identity, runtime protection, inference infrastructure, autonomous workflows, and third-party risk. The firm expects cybersecurity to become an increasingly important enabler of enterprise AI deployment rather than simply serving as a compliance function.

Fund continues firm’s infrastructure-first investment approach

Jump Capital said the new vehicle represents a “picks-and-shovels” approach to AI adoption, with planned investments across vertical AI applications, cybersecurity, and the infrastructure supporting enterprise deployment.

The firm said it will continue writing initial checks ranging from $1 million to $4 million and larger investments between $8 million and $15 million. It has also expanded its presence in New York over the past two years while continuing to invest across North America and Israel, where it has developed relationships with technical founders.

The Information, citing Jump Capital co-founder and partner Sach Chitnis, reported that the firm’s recent AI investments include compensation software company Compa, GPU software automation startup Standard Kernel, and AI infrastructure platform TrueFoundry.

Previous investments show a focus on foundational technologies

Although the new fund centers on artificial intelligence, Jump Capital’s broader investment history includes backing foundational technologies across multiple sectors.

In May 2025, Jump Crypto, the digital asset investment division that operates separately from Jump Capital, acquired a significant equity stake in real-world asset tokenization platform Securitize. The company said the partnership would help expand institutional access to tokenized Treasurys, private credit, and private equity while improving collateral management.

A month later, Jump Crypto partnered with Aptos Labs to introduce Shelby, a decentralized storage network designed for data-intensive Web3 applications. The project was built to provide cloud-grade decentralized storage with sub-second reads while addressing blockchain limitations around storing and serving large datasets.

The firm’s blockchain investment activity continued in September 2025, when Jump Crypto participated in KGeN’s $13.5 million strategic funding round alongside Accel and Prosus Ventures. KGeN said the capital would expand its on-chain identity and reputation framework supporting user acquisition, commerce, and loyalty programs, while Jump Crypto Chief Investment Officer Saurabh Sharma said the platform introduced greater accountability to digital distribution.

Technical founders remain at the center of Fund VIII

Jump Capital said the pace of AI development has made it increasingly difficult to distinguish durable businesses from short-term opportunities, with stronger competition for talent and higher expectations for new products.

The firm said it continues to favor technical, product-oriented founders who rethink business workflows around AI rather than simply using the technology to improve existing processes. It added that many emerging companies are designing organizations around software-first execution while carefully identifying where human expertise continues to add value.

Looking ahead, Jump Capital said Fund VIII will continue supporting founders modernizing legacy industries, building enterprise AI infrastructure, and securing autonomous software systems as AI adoption accelerates across businesses.

BitGo Holdings launched BitGo Link on Aug. 3, giving institutional trading and treasury teams one interface for viewing and moving capital held across BitGo and connected cryptocurrency exchanges.

The New York Stock Exchange listed digital asset infrastructure provider described Link as a centralized control layer for external exchange accounts. It allows clients to monitor buying power, initiate transfers and manage user permissions without signing into each venue separately.

The product connects with BitGo’s wider custody, settlement and prime brokerage services. However, BitGo did not disclose Link’s fees, customer numbers, supported assets or a complete list of connected exchanges in its official announcement. It said only that the product connects clients with a “large network” of leading venues.

BitGo Link centralizes balances and exchange transfers

Link gives institutions a consolidated view of assets held inside BitGo and in accounts at outside exchanges. The dashboard includes subaccounts and calculates the client’s total buying power across connected venues, according to the company’s official product page.

Trading teams can use the platform to transfer funds between accounts, rebalance capital or meet margin and liquidity requirements. BitGo says Link automatically compares transfers with exchange records and displays each transaction’s progress from initiation through settlement.

The company presents the service as an answer to a common institutional problem. Firms may hold assets with a custodian while maintaining separate accounts at several exchanges. Each venue can have different permissions, reporting systems and settlement processes.

Link does not mean that every connected asset remains in BitGo custody. The service provides visibility and transfer controls for both BitGo and non BitGo accounts. That differs from the company’s off exchange settlement service, where eligible assets can remain in segregated custody while an exchange receives a projected balance for trading.

Policy controls follow capital to connected venues

Every transfer initiated through Link passes through BitGo’s Policy Engine. Institutions can apply approval requirements similar to those already used for transfers from BitGo wallets, including role based access and internal authorization rules.

Administrators can also extend permissions across connected exchange accounts from one place. BitGo said this removes the need to recreate access settings at each venue, although clients will still depend on the technical availability and terms of the participating exchanges.

CEO Mike Belshe described Link as BitGo’s “command center for institutional treasury and trading.” That language reflects the company’s positioning of the product. BitGo has not released usage data showing how many institutions have adopted it or how much capital it currently manages.

The company also identified technical integration failures, transaction errors, digital asset volatility and regulatory scrutiny among the risks that could affect its broader products. Those disclosures mean Link’s planned expansion should be treated as a company objective rather than a confirmed outcome.

Link extends BitGo’s institutional connectivity push

Link complements Go Network, which provides settlement and access to exchange liquidity within BitGo’s custody framework. It also connects with BitGo Prime, which offers execution, financing, collateral management and access to liquidity providers.

As crypto.news reported in related coverage, Gate US joined Go Network in July. That arrangement lets eligible institutional clients trade against the exchange’s liquidity while supported assets remain in BitGo Bank & Trust custody until settlement.

Link addresses a different part of the operating process. It gives treasury teams control over assets already distributed among external accounts, rather than limiting the workflow to balances held under the company’s qualified custody structure.

BitGo has also widened the services available from custody. As previously reported, the firm opened controlled access to Aave, Spark and Tesseract in June. Narval software checks approved contracts and policy rules before BitGo authorizes transactions from eligible custody wallets.

The company said it plans to connect Link with more exchanges and treasury processes over time. It did not provide a rollout calendar or identify the next venues under consideration.

Investors await BitGo’s August 12 earnings report

BTGO last traded near $5.06 on Aug. 3, approximately 4.1% above its previous close of $4.86. The available trading data does not establish that the Link announcement caused the increase.

The share price remained well below BitGo’s $18 January IPO price. The company raised approximately $174.3 million in net proceeds through the offering, according to its first quarter filing.

BitGo reported $3.77 billion in first quarter revenue, up from $1.77 billion one year earlier. Its net loss widened from $25.7 million to $60.7 million. The loss included unrealized changes in the value of its digital asset treasury and expenses connected with the IPO.

Most of the reported revenue came from digital asset sales recorded on a gross basis. Digital asset sales generated $3.66 billion, with associated costs of $3.65 billion. The accounting presentation means the headline revenue figure should not be read as an equivalent measure of service fees or operating profit.

The launch also follows a company restructuring. As crypto.news previously reported, BitGo cut nearly 15% of its workforce in June while directing resources toward security, trading, stablecoins, settlement and artificial intelligence infrastructure.

BitGo will publish second quarter results after the market closes on Aug. 12. Management will hold an earnings call at 5 p.m. Eastern Time. The report may provide the next verified update on client growth, costs and whether newer institutional products are contributing to the business.

Galaxy Research has estimated that losses tied to the Coldcard hardware wallet vulnerability have reached 1,596 Bitcoin across confirmed attack waves and could climb to about 2,055 BTC, or nearly $130 million, if a suspected fourth wave is verified.

Galaxy Research said in a post published Monday on X that it has identified 1,596 BTC stolen from 7,300 addresses across three confirmed waves of attacks, along with 14 smaller security incidents linked to the Coldcard seed-generation flaw.

The research firm said its latest estimate excludes a fourth suspected attack wave because it has not yet received enough confirmation from affected wallet owners. If the additional activity is validated, the total would rise to 2,055 BTC, valued at about $130 million at current prices.

Earlier blockchain analysis from Galaxy had estimated roughly 1,815.75 BTC moving across four observed waves, but the firm emphasized at the time that the figures came from on-chain analysis rather than confirmed victim reports. The latest update narrows confirmed losses while keeping the larger estimate tied to the still-unverified fourth wave.

Coldcard attack investigation remains active

Galaxy said it has identified what it believes is a fourth coordinated wave of theft but is still waiting for victim confirmation before adding those addresses to its confirmed tally.

According to the research firm, blockchain activity suggests the suspected fourth wave is “substantially comprised of” one attacker, giving analysts medium-high confidence in that assessment despite the remaining uncertainty over affected wallets.

Alex Thorn, Galaxy’s head of firmwide research, first flagged the potential fourth wave on Aug. 3 after identifying transaction patterns that closely matched the earlier attacks. His running estimate later rose to 448.7 BTC moving from 709 potential victim addresses, although Galaxy stressed that blockchain data alone cannot confirm every victim or determine whether a single operator carried out every theft.

The firm added that its investigators continue refining address mapping as additional information becomes available from wallet owners and other participants in the investigation.

Coldcard flaw dates back to 2021 firmware change

The attacks stem from a vulnerability affecting seeds generated on Coldcard Mk3, Mk4, Mk5 and Coldcard Q devices running vulnerable firmware versions.

Coinkite disclosed last week that the flaw originated in March 2021 while integrating a new cryptographic library into its firmware. Instead of generating wallet seeds through the intended hardware-backed true random number generator, affected firmware mistakenly relied on a deterministic pseudo-random generator provided by MicroPython.

According to Coinkite’s technical review, the hardware random-number generator remained active elsewhere in the firmware, allowing internal reviews to confirm its presence without revealing that wallet creation had switched to a different entropy source.

Block’s Bitcoin engineering and security team independently reached the same conclusion after reviewing the firmware. While the company said it had not completed full empirical testing of every affected device, it concluded that the vulnerable firmware called the deterministic MicroPython fallback instead of the STM32 hardware random-number generator during seed creation.

Coinkite estimates that affected Mk2 and Mk3 devices may provide roughly 40 bits of effective entropy, while vulnerable Mk4, Mk5 and Coldcard Q models may generate about 72 bits instead of the intended 128 bits.

Most stolen Bitcoin has not moved

Galaxy said investigators have been working with U.S. federal law enforcement agencies, cryptocurrency exchanges and cyber investigation groups by sharing confirmed attacker and victim addresses as the investigation expands.

The research firm reported that approximately 90% of the stolen Bitcoin remains untouched. It added that none of the coins stolen during the first three confirmed attack waves have moved since they were taken, giving investigators additional time to monitor the funds.

Galaxy also warned that new attackers could attempt to exploit the same vulnerability while affected devices remain in use. For that reason, it said identifying additional attacker-controlled addresses remains important so exchanges and authorities can respond if funds begin moving.

Earlier blockchain analysis showed attack activity accelerating to about 13.8 wallet sweeps per Bitcoin block during the fourth suspected wave, compared with roughly 0.3 sweeps per block before the incident. Galaxy also observed that most stolen balances were transferred to newly created addresses instead of one central collection wallet, while some funds later moved through second-hop transactions that complicated blockchain tracing.

The firm previously noted that users who still control compromised wallets may have a limited opportunity to replace an unconfirmed theft transaction with a higher-fee transaction under Bitcoin’s Replace-by-Fee mechanism, although the option only exists before miners confirm the original transaction and offers no guarantee of recovery.

Coldcard users are still urged to replace wallet seeds

Galaxy said the attacks remain active and advised affected Coldcard users to move their funds to secure addresses and create entirely new wallet seeds on patched devices.

Coinkite has already released emergency firmware updates for every affected product, including version 4.2.0 for Mk2 and Mk3 devices, version 5.6.0 for Mk4 and Mk5, version 1.5.0Q for Coldcard Q, and Edge releases 6.6.0X and 6.6.0QX. The company has also destroyed all remaining inventory containing vulnerable firmware.

According to Coinkite, installing updated firmware protects only wallets created after the fix. Existing seed phrases generated with vulnerable firmware remain exposed and should be replaced.

The company recommends generating a completely new seed after updating the device, verifying a receiving address, sending a small test transaction, and transferring the remaining balance only after confirming the test succeeds.

Coinkite also said wallets created using at least 50 fair private dice rolls are not exposed by this random-number-generation issue alone. While a strong BIP-39 passphrase adds another security layer, the company continues to recommend migration because the original vulnerable seed remains weak.

Tyler Williams, one of Treasury Secretary Scott Bessent’s senior digital asset advisers, left the U.S. Treasury Department on July 31 as Congress struggled to advance the CLARITY Act before its August recess.

Bessent confirmed the departure in a statement reported by Punchbowl News on Aug. 3. He called Williams “instrumental” in advancing the administration’s goal of making the United States the “crypto capital of the world.”

Punchbowl reported that Williams is “expected to return to the private sector,” although neither his next employer nor his future position was disclosed.

XRP holders can now borrow against their coins on Ethereum without selling them.

Flare, the blockchain that lets XRP be used in decentralized finance, told CoinDesk on Monday that its wrapped version of the token has been accepted as collateral by Sentora, which manages a $280 million lending pool of Ripple’s RLUSD stablecoin.

Borrowers post the wrapped XRP, called FXRP, and take out RLUSD against it.

The approval opens an isolated FXRP/RLUSD market on Morpho Blue where holders can borrow against XRP exposure rather than sell it. Access is permissionless with no whitelist required.

Sentora reviewed FXRP’s market behavior, oracle design, liquidity and liquidation capacity under its institutional risk framework before signing off, and the asset will face the same ongoing monitoring as other collateral in the vault.

Getting there takes several steps today. Users mint FXRP through Flare’s FAssets system, bridge it to Ethereum via Stargate, deposit it into the market and borrow at their chosen loan-to-value ratio. Flare says it is building a route through Smart Accounts that would let holders authorize the whole sequence from an XRP Ledger wallet, with direct XRPL-to-Ethereum minting also in development.

Crypto’s biggest players are still waiting to gain access to powerful new AI models to strengthen their code from attacks, but only a select few have been able to get it. 

US crypto exchange Coinbase said in June that it had secured access to Anthropic’s restricted Mythos model and Zcash’s Zooko Wilcox said Anthropic used the model to audit the Zcash protocol at the request of Shielded Labs, while other major crypto players are seemingly yet to get access.

“That’s one advanced frontier model that hasn’t been made available to crypto just yet,” Binance’s chief security officer Jimmy Su told Cointelegraph. “We have been trying to make inroads there. We also talked to other crypto exchanges and our own investors to try to make some progress. But we haven’t gotten the most frontier AI model, like Mythos.” 

The uneven access creates a new security divide in an industry where exploits can put billions of dollars at risk. While model developers like Anthropic and OpenAI have chosen to restrict their most cyber-capable models from the public, there are concerns that increasingly powerful open-source alternatives mean crypto firms are left to deal with sophisticated AI-assisted attacks without the most capable tools to defend against them. 

Crypto executives say restricted access is initially necessary

Anthropic says Mythos 5 uses the same underlying model as its publicly available Fable 5, but without safeguards that restrict sensitive cybersecurity work. 

OpenAI operates a similar tiered system: verified defenders can use GPT-5.5 with “Trusted Access for Cyber”, while its more permissive GPT-5.5-Cyber model is reserved for a smaller group conducting authorized penetration testing.

Crypto security executives interviewed by Cointelegraph said there is likely a need to initially restrict access to frontier cyber models, but said continuing to gate them becomes harder to justify once publicly available models approach the same capabilities. 

Source: Zooko Wilcox

Su said Anthropic’s controlled rollout is a responsible approach because newly released models may benefit attackers faster than defenders. 

“If it enhances the attacker much faster than the defender, then it actually is harming the ecosystem,” he said, adding that a limited testing period could reduce the potential “blast radius.”

Related: Can AI drain DeFi? Separating Claude Mythos hype from reality

However, Su said this calculation changes when competing models become more powerful and widely available. 

“As other more powerful models are being released, the pressure will be on Anthropic to make it more widely available,” he said. The question would be whether defenders can deploy the frontier model as effectively as attackers once it becomes available, he said. 

The number of critical-severity CVEs has climbed after the launch of Claude Mythos Preview. Source: Epoch AI

Solana Foundation chief information security officer Michael Coates, who joined the foundation in July, also supported safeguards but argued that legitimate defenders need a faster route to them. 

“I fully understand guardrails for advanced models, but we need to streamline the verification programs, the acceptance programs, to give these models to legitimate defenders,” he said.

“We need to make sure that the best models we can get are in the hands of defenders because attackers will have something capable enough.”

Blockchain Capital’s Sean Cheetham also supported eventually opening up restrictions, and said that broader availability could ultimately favor defenders as legitimate security researchers greatly outnumber the small groups conducting sophisticated attacks. 

“If good people can multiply their defense scale… you’re much better off just opening it up and allowing them to defend themselves,” he said.

Uneven access to frontier AI models 

Binance’s lack of access comes despite it being the biggest crypto exchange in the world by daily trading volume. The exchange holds a total of $137.8 billion in assets, according to DefiLlama. 

Crypto custodian Fireblocks, which secures trillions in assets annually, said in April it has sought access to Mythos and at the time, only used Anthropic’s publicly available model for pentesting, according to The Information, while Uniswap founder Hayden Adams in June slammed Fable 5’s safeguards that restrict prompts relating to cybersecurity. 

The Ethereum Foundation in July said it has been running “coordinated AI agents” to find bugs across its systems, but didn’t disclose which models were being used. 

Cointelegraph reached out to Ethereum Foundation, Fireblocks and Uniswap to confirm if they have since received access to frontier AI models. 

Source: Hayden Davis

Meanwhile, some crypto-adjacent companies have gained access. FIS, which provides technology to banks and partnered with Circle in July last year to let banking clients offer domestic and cross-border payments in USDC, joined Project Glasswing last month. 

Project Glasswing is Anthropic’s gated program for giving vetted cyber defenders and organizations responsible for critical software infrastructure early access to its restricted Mythos models.

HackerOne, which provides bug-bounty and security testing services to major crypto exchanges, among others, also said it joined Project Glasswing, though testing is confined to its own infrastructure, not its customers’ programs. 

Cointelegraph reached out to OpenAI and Anthropic about how many crypto companies have been given access to restricted models. 

AI-assisted hacking attempts on the rise

On Monday, Bitcoin swap service Boltz said it has chosen to halt its non-custodial bridge after seeing a steady rise in AI-assisted exploits over the past few months. 

“The pattern is clear: attackers now iterate faster than a team our size can find and patch.”  

Last week, Bitcoin hardware wallet company Coinkite said a number of its Coldcard devices were exploited due to a flaw in its wallet seed generation, which turned out to be less random than expected. 

It speculated that the attacker had used AI to review previous versions of the firmware to find and exploit the flaw, despite it using “one of the best available AI models” to review its code just weeks before. 

Magazine: Fears of AI-driven DeFi hack epidemic overstated for now — but not for long

Boltz suspended its Bitcoin swap services indefinitely on Aug. 3 after reporting months of automated, AI assisted probing and several contained exploits. 

The noncustodial swap provider said the attacks had accelerated sharply in recent days, leaving its development team unable to deploy fixes as quickly as attackers adapted their methods.

The team said it could not responsibly restore swaps while it remained under active attack and reviewed findings from recent security scans. Boltz described the suspected attackers as “multiple resourceful groups,” although it did not identify them or provide evidence independently confirming who conducted the attacks.

Boltz swap shutdown followed an earlier EVM warning

The wider shutdown followed an Aug. 1 notice concerning a bug in Boltz’s Ethereum Virtual Machine integration. The service initially disabled swaps involving assets such as USDT, USDC, WBTC, TBTC and RBTC while stating that Bitcoin, Lightning and Liquid swaps remained operational. Two days later, it stopped all swap services.

Boltz operates infrastructure connecting Bitcoin mainnet with the Lightning Network, Liquid and other supported networks. Its public API powers the main web application and outside integrations. Official documentation advises developers to use supported software development kits because they manage swap cryptography, recovery procedures and transaction states.

Moreover, Boltz said no customer funds were exposed during the incidents because users retain control of their assets throughout its atomic swap process. The company also said it absorbed the losses associated with the contained exploits because it operates as a fully bootstrapped business. Those statements remain company claims because Boltz has not published a technical incident report or independent security review.

The service’s API remains online for cooperative refunds. Users can also complete unilateral refunds without relying on Boltz infrastructure, according to the company. Its support team remains available for customers with unfinished transactions.

The distinction shows how noncustodial architecture can limit custody losses without preventing operational disruption. Users may retain their Bitcoin, but they cannot initiate new swaps through the affected service until operations resume or integrated wallets introduce alternative providers.

Wallets are seeking replacements for Boltz infrastructure

Bull Bitcoin said the shutdown temporarily disabled Lightning payments and conversions between Liquid Bitcoin and onchain Bitcoin within its wallet. Standard Bitcoin transfers, Liquid transfers, wallet restoration and storage functions continued operating normally.

The company said it was evaluating several replacement options and promised a separate mechanism for customers wishing to convert Liquid Bitcoin into onchain Bitcoin. It added that its Liquid Federation membership allows it to conduct conversions without depending on a third party.

ZEUS also disabled its own deployment of the open source Boltz stack, while Aqua notified users that the service suspension affected its swap functions. Neither wallet reported customer asset losses.

As crypto.news previously reported, Blockstream had integrated Boltz into its mobile wallet to support Lightning and Liquid swaps. The shutdown shows how applications that use shared swap infrastructure can face service interruptions even when their core wallet functions remain available.

Boltz has not provided a reopening date

Boltz warned users not to expect swap services to resume soon. It has not published a remediation timetable, detailed vulnerability list or conditions that must be met before operations restart. The team said it would provide another update after assessing its options.

The company’s attribution to AI assisted activity also remains difficult to verify without technical indicators. Automated scanning can increase the speed and volume of attacks, but Boltz has not explained how it determined artificial intelligence played a role.

In related coverage, crypto.news reported that Solana Foundation security chief Michael Coates expects artificial intelligence to strengthen both attackers and defensive systems. Boltz’s next update is expected to clarify whether it will introduce automated monitoring, outside audits or changes to its open source infrastructure before restoring swaps.