More than a dozen bitcoin companies and nonprofits (including Coinbase, Block, BitGo and Blockstream) are asking the leading AI labs to give their security researchers the same access to frontier models that attackers already exploit, CoinDesk's Shaurya Malwa reported.
Organized by the Bitcoin Policy Institute, the letter makes five requests: early access to models before public release, enough compute to run them, secure environments for reviewing private code, eligibility for small independent maintainers, and a direct line to the labs' security teams. The sticking point, the signatories say, is that Bitcoin Core developers "cannot get into the programs labs run for trusted security partners," while consumer models' safety filters wave off legitimate security work.
The ask lands as AI collapses the gap between finding a bug and exploiting it. The volunteer Bitcoin Red Team scanned open-source bitcoin code with AI agents this month and logged thousands of findings, and recent exploits of BTCPay Server and Foundation's hardware-wallet node show the stakes.
The DeAI takeaway: The asymmetry the letter describes is, at bottom, a closed-model problem, one that exists because the most capable models sit behind a handful of labs' access programs. Open-weight models carry no such gatekeeper — a contrast worth watching as the defense of open-source bitcoin increasingly hinges on who those labs choose to let in.