Crypto coalition flags AI security gaps

TLDR

Over 40 Bitcoin and crypto organizations say AI labs are unintentionally giving attackers a head start over the teams defending digital-asset infrastructure.

1. A broad industry coalition signed an open letter asking frontier AI labs to grant vetted security defenders controlled early access to their strongest models.
2. They argue attackers already use powerful AI to find wallet and node bugs faster, while maintainers are stuck with weaker tools and flooded with low-quality AI bug reports.
3. The next key signals are how AI labs respond, whether regulators join the discussion, and whether security programs adapt to AI-driven exploit and phishing trends.

Deep Dive

1. Coalitions Core Message

Groups including Coinbase, Block, BitGo, Blockstream, ARK Invest, Anchorage Digital and others backed an open letter titled Defenders Need the Frontier, coordinated by the Bitcoin Policy Institute. The letter warns that frontier AI systems are accelerating vulnerability discovery in software that secures Bitcoins trillion?dollar stack and other digital assets, but most open-source maintainers cannot access these models under current lab policies, creating a dangerous security divide.

The signatories ask labs to create trusted-access programs for qualified defenders: early access to cyber-capable models, enough compute to scan large codebases, secure environments for reviewing private code, inclusion of small nonprofits and volunteers, and direct channels to lab security teams. The request focuses on defensive use, not wider public release of the most capable models, as described in the coalitions open letter.

2. How AI Is Shifting Crypto Security

Recent incidents show both the upside and downside of AI for crypto. A volunteer Bitcoin Red Team used AI models to audit 390 projects in 27.5 hours, producing 4,962 findings, including critical issues in BTCPay Server that allowed Lightning nodes to be drained, highlighting AIs power as a defensive accelerator.

At the same time, hardware wallets like Coldcard and other services have faced suspected AI-assisted exploit discovery, and Coinbase reports its bug bounty submissions could triple in 2026 as AI tools churn out large volumes of low-quality reports, with only 4 percent of HackerOne submissions paying out, according to a Coinbase security update. Onchain, one assessment counted 212 exploits and 1.1 billion dollars in losses in the first half of 2026, with privileged key misuse dominating losses.

3. What To Watch Next

So far, major AI labs have not publicly committed to the specific access changes requested, even as they work on stronger safeguards against malicious use of their models. Any announcement of defender-focused access programs, compute grants, or dedicated security model tiers would be important signals.

Regulators are also beginning to pair crypto and AI in oversight discussions; for example, the CFTCs Innovation Advisory Committee plans to discuss both technologies in its inaugural meeting, indicating that AI security will increasingly be framed as a market-integrity issue. For individual users and firms, the practical implication is to assume attackers can and do use AI, and to treat secure key management, careful software choice, and up-to-date clients as baseline rather than optional.

Conclusion

Crypto security is entering an AI-heavy phase where code audits, phishing, and exploit development are all accelerated by machine intelligence. The coalitions warning is simple: if defenders cannot access similarly powerful tools under safe, controlled programs, open-source financial infrastructure may lag attackers. How AI labs and regulators respond will shape whether AI becomes a net defensive advantage or a persistent structural risk for digital assets.