Coinbase and 40 Firms Demand AI Access to Defend Bitcoin

More than 40 cryptocurrency companies and developer groups, coordinated by the Bitcoin Policy Institute, signed an open letter on August 10 asking frontier AI laboratories to give vetted Bitcoin security researchers controlled access to their most capable models. The letter, titled "Defenders Need the Frontier," says the infrastructure in question protects over $1 trillion in value and argues that public AI safety restrictions are blocking legitimate cybersecurity research while open-source maintainers are left relying on weaker open-weight models.

Signatories include Block, Coinbase, Strategy, MARA, Galaxy, BitGo, Blockstream, Anchorage Digital, ARK Invest, Bitwise, Kraken, Ledger, Foundry, Trezor, and BTCPay Server, plus developer groups such as Brink, Chaincode Labs, OpenSats, and Btrust. The coalition is asking for early access to frontier cybersecurity models, dedicated computing capacity, a secure research environment, and direct channels with AI lab security teams.

Rising incident numbers

The push comes as security incident data points to a sharp rise in attack frequency. SlowMist recorded 182 blockchain security incidents in the first half of 2026, resulting in roughly $956 million in losses. That compares with 121 incidents and $2.37 billion in losses in the same period of 2025. Incident frequency rose about 50% year over year, while the average loss per event fell sharply. TRM Labs counted 207 incidents and $972 million in losses for the same period, pointing in the same direction.

The letter follows recent examples of AI-assisted security work. The Bitcoin Red Team, a volunteer group led by a contributor known as Calle, used AI models plus manual review to scan open-source repositories. On August 13, Calle posted on X that the effort surfaced revelations, calling it "a massive collision between decades of human open source slop against 2 weeks of kimi k3," a reference to Moonshot's AI model. Separately, an Ethereum Foundation study in July used coordinated AI agents to find actual vulnerabilities in Ethereum software, including a libp2p flaw published as CVE-2026-34219.

The coalition is not asking for unrestricted public access. Instead, it proposes repeated trusted-access programs for qualified defenders of open-source financial infrastructure, with researchers vetted before receiving capabilities withheld from the general public. No leading AI laboratory had publicly announced a program responding to the requests at the time of publication.