Crypto firms warn on AI security gap

TLDR

More than 40 crypto and Bitcoin firms have warned that current AI rules leave security defenders using weaker tools than the attackers trying to hack them.

1. An industry coalition led by the Bitcoin Policy Institute published an open letter urging frontier AI labs to give vetted security researchers early access to powerful models.
2. The gap is already visible in practice, with AI accelerating both bug discovery in Bitcoin infrastructure and sophisticated phishing and exploit campaigns against crypto users.
3. The firms propose trusted-access programs for white-hat defenders, and the next key signal will be how major AI labs respond and what protections they attach.

Deep Dive

1. Who Is Warning And About What

Over 40 Bitcoin and digital-asset organizations, including Coinbase, Block, BitGo, Blockstream, Anchorage Digital, ARK Invest, Bitwise, developer funds Brink and Chaincode, and others, signed an Aug. 10 open letter titled Defenders Need the Frontier, published by the Bitcoin Policy Institute.

The letter argues that maintainers of open-source financial infrastructure, such as Bitcoin Core and wallet projects, are blocked by safety guardrails and vetting programs on leading AI models, while attackers face few constraints and can already use frontier systems for vulnerability hunting.

Reports from outlets like Bitcoin.com and Decrypt describe this as a growing security divide, where the software securing more than a trillion dollars in Bitcoin and related assets is audited with weaker AI than the tools criminals can access.

2. How The AI Security Gap Shows Up Today

Recent incidents highlight both sides of AIs role in crypto security. A volunteer Bitcoin Red Team used advanced models such as Moonshots Kimi K3 to scan 390 Bitcoin projects, generating thousands of findings, including dozens of critical issues, showing AIs defensive power.

At the same time, a critical flaw in BTCPay Servers Lightning integration was exploited to drain merchant nodes, with the project itself noting that AI is making it cheaper and faster to search large codebases for weaknesses. Separate reporting on the Zoomsday Zoom exploit shows a researcher using fewer than 20 AI prompts to build a zero-click attack that could expose wallets and exchange sessions.

Coinbase has also warned that AI is flooding its bug bounty program with low-quality reports, tripling submission volume while valid bugs fall as a share of total, even as criminals use AI to supercharge phishing and credential theft campaigns.

3. What Crypto Firms Are Asking For And What To Watch

The signatories are not asking for unrestricted public release of powerful models. Instead, they want AI labs to create or expand trusted-access programs that give qualified defenders: early access to the strongest cyber-capable models, enough compute to run serious reviews, secure environments for private code, eligibility for small nonprofits and independent maintainers, and direct channels to lab security teams.

So far, major labs have not issued detailed public responses on this specific request, though they acknowledge that frontier models can both help defenders and make ransomware and other crypto-related attacks cheaper. The letter remains open for additional signatures, and further Bitcoin Red Team disclosures are expected.

Conclusion

Crypto firms are effectively telling AI labs that security is a parity problem: if attackers can use frontier AI, defenders need comparable tools or the balance tilts toward more and faster exploits.

How labs respond to this open letter, and whether they open structured access for white-hat crypto security teams, will shape the resilience of Bitcoin and broader digital-asset infrastructure in the next wave of AI-driven attacks.