More than three dozen bitcoin and crypto organizations are asking major AI labs to give open-source security researchers early access to their strongest cyber-capable models. The letter was organized by the Bitcoin Policy Institute, with signatories cited by CoinDesk including Coinbase, Block, BitGo, Blockstream, Anchorage Digital, ARK Invest, Bitwise, Foundry, Casa and Exodus, alongside developer funds such as Brink, Chaincode and Btrust.
Their central argument is an asymmetry between defenders and attackers. Bitcoin Core developers and independent maintainers often do not qualify for trusted-partner programs run by AI labs. Public models, meanwhile, may block prompts that resemble malicious code, which can interfere with legitimate vulnerability research. Attackers are not limited in the same way if they use open-weight models, stolen access or purpose-built offensive tools.
The signatories are asking for five concrete resources: pre-release access to the most capable cyber models, enough compute budget to conduct meaningful reviews, secure environments for private code, eligibility for small independent maintainers, and a direct reporting channel to lab security teams. Recent incidents affecting BTCPay Server and Lightning nodes are cited as evidence that the threat is already operational rather than theoretical.
A volunteer Bitcoin Red Team provides another example. The group has started applying AI models across bitcoin codebases and has filed thousands of findings across hundreds of projects; one report contributed to a BTCPay fix. The request is therefore not simply to remove AI safeguards, but to create a trusted path where defenders of critical open-source infrastructure can use comparable capabilities before new offensive techniques become widely available.