Bitcoin firms seek AI security access to find flaws before hackers

Bitcoin firms AI security access is the focus of a new request from more than 40 Bitcoin and crypto companies, which have asked major AI labs to let vetted independent researchers use their strongest models before those tools are publicly released. For investors, the issue is less about an AI attack on Bitcoin itself and more about the wallets, custody providers and software that connect people to crypto assets.

According to Yahoo Tech, the group published its letter on Aug. 10. It seeks early access and reasonable computing capacity for qualified defenders of open-source financial infrastructure through ongoing trusted-access programs.

The companies argue that public AI systems can refuse legitimate cybersecurity requests because of safeguards designed to prevent misuse. In the letter's view, that can leave the people responsible for finding and fixing flaws with weaker tools than attackers may obtain.

What are Bitcoin firms asking AI labs to provide?

The request is for controlled access, not the broad public release of more powerful hacking-capable models. Researchers who meet a lab's qualification standards would be able to use pre-release or advanced systems to test code, identify vulnerabilities and help developers fix problems before criminals exploit them, Yahoo Tech reported.

That distinction matters because the signers are describing an access policy for defensive work. AI labs would still have to decide who is trusted, how activity is monitored and how to limit the risk that a model or its findings are misused.

The risk is largely around crypto's surrounding systems

The letter's rationale covers the broader software stack around Bitcoin, including wallets, signing devices, code libraries and custody systems. Yahoo Tech reported that the signers said the Bitcoin network secures more than $1 trillion, while the surrounding infrastructure holds trillions more. Those figures are the group’s stated case for giving defenders stronger tools.

Bitcoin's blockchain and the businesses built around it face different risks. CNBC reported that the Bitcoin blockchain has operated without interruption since 2009, while Yan Pritzker, chief technology officer at Swan Bitcoin, said the network's cryptography is not affected by AI and its rules require broad consensus to change.

Exchanges and other platforms handling customer funds and personal information may present more immediate targets. Cosmo Jiang, a general partner at Pantera Capital, told CNBC that financial-services companies and exchanges would likely be targeted first, though the risk is wider. Clear Street analyst Owen Lau told CNBC that AI agents could help create high volumes of scam emails and synthetic identities.

AI can help attackers and defenders

The same capability that speeds up a search for software flaws can also help security teams test systems more thoroughly. Coinbase Chief Security Officer Philip Martin told CNBC that future models could enable deeper testing at scale, accelerating both digital threats and defenses. Binance Chief Security Officer Jimmy Su said the company was experimenting with AI to identify vulnerabilities faster and across more of its systems.

Major AI labs had not publicly committed to the firms' specific proposal as of Yahoo Tech's report. The request also differs from a separate $15 million commitment made in late July by the Bitcoin Security Consortium toward long-term Bitcoin-network security, according to Yahoo Tech.

This story draws on original reporting from Decrypt.