Bitcoin security researchers say U.S. AI restrictions are driving defensive work to Chinese open models

Bitcoin security researchers and company leaders say restrictions on U.S. frontier AI systems are forcing defenders of more than $1 trillion in crypto infrastructure to rely on Chinese open-source models that will perform legitimate code audits. The pressure coincided with an August 10 open letter from the Bitcoin Policy Institute signed by more than 40 Bitcoin and digital-asset organizations, including Coinbase, BitGo, Block, Blockstream and Anchorage Digital, calling on major AI labs to provide vetted open-source security researchers with controlled early access rather than unrestricted public release of highly cyber-capable models. The signatories said wallets, cryptographic libraries, payment processors, exchanges, custody services and Lightning Network tools across the Bitcoin ecosystem depend heavily on open-source software, while attackers are already using AI to accelerate vulnerability discovery and defenders are often constrained by model safety policies. They proposed trusted-access programs that would include pre-release models, compute resources, private code-review environments and direct communication channels with lab security teams. The appeal came after the volunteer Bitcoin Red Team used AI tools to audit Bitcoin-related code earlier in August, reviewing 390 projects in about 27.5 hours and submitting roughly 4,962 findings, including dozens of critical issues and hundreds of high-severity issues. Participants said AI-generated findings still require human verification to confirm whether vulnerabilities are real and to assess their severity. The risks are not theoretical: BTCPay Server previously disclosed that attackers actively exploited a flaw in older versions that could let them obtain LND administrator credentials and take control of associated wallets.

The information on this website is generated using AI and we cannot guarantee its accuracy. Please use it as reference information only.