Anthropic’s AI coding assistant Claude Code has reportedly identified the vulnerability behind a Bitcoin hardware wallet exploit that caused losses estimated at more than $100 million. The claim surfaced after BitGo CEO Mike Belshe publicly challenged Claude to steal 100 BTC from a BitGo wallet, claiming that practical evidence is more important than hypothetical worries about AI-powered hacking. The two advancements have rekindled discussion about AI’s expanding role in cybersecurity and crypto security.
According to developers and security researchers, Claude Code was asked to inspect the source code of the Coldcard hardware wallet for security flaws. After roughly eight minutes of analysis, the AI reportedly identified the same vulnerability that attackers had previously exploited.
Researcher Medusa highlighted the result on X, saying Claude found the flaw with a single prompt after just eight minutes of reasoning. The wallet’s random number generation procedure, which is essential for generating safe private keys and digital signatures, was connected to the vulnerability. Attackers could be able to retrieve wallet keys and steal money if randomness becomes predictable.
The exploit has been described as one of the largest hardware wallet compromises in Bitcoin history. Early estimates suggested more than 1,080 BTC was stolen in less than an hour, while more recent community estimates place total losses above $100 million. Security experts also believe the attacks may not be over. Galaxy’s Head of Research, Alex Thorn, recently pointed to a fresh wave of suspicious on-chain transactions affecting hundreds of Coldcard wallet addresses and urged users to move their funds immediately.
Either @AnthropicAI is terrible at building sandboxes… or excellent at marketing. (or both)
But enough with the “we created a hacking monster” games.
Do it for real.
I put this in an @BitGo wallet for you. Go get it.
100 BTC:… https://t.co/RhvivRk9YK
— Mike Belshe (@mikebelshe) August 1, 2026
The discussion around AI security intensified further after BitGo CEO Mike Belshe issued a public challenge to Anthropic. On August 1, he revealed a Bitcoin wallet holding 100 BTC and invited Claude to “go get it,” saying real demonstrations would carry more weight than claims that advanced AI models have become dangerous hackers.
The issue arose when Anthropic revealed that operational errors had caused some of their Claude models to interact with actual computer systems during cybersecurity assessments. The business claims that rather than any intentional attempt by the AI to break out of its sandbox, the instances were caused by testing settings inadvertently staying online.
Security experts pointed out that Belshe’s task is far from simple despite the fanfare. The secret keys required to transfer money are not disclosed by publishing a Bitcoin address, and transactions involving BitGo’s multisignature wallets usually need many separate approvals. Therefore, rather than cracking Bitcoin’s encryption, effectively stealing the cryptocurrency would need jeopardizing BitGo’s operational security.
The episode demonstrates how quickly AI is being discussed in relation to cybersecurity. Although current AI models are unable to decipher Bitcoin’s encryption, experts caution that they may progressively assist hackers in identifying software vulnerabilities, taking advantage of lax security procedures, and automating complex cyberattacks. Although Belshe’s 100 BTC challenge is still unresolved, it has spurred new discussion on how the cryptocurrency sector should get ready for more powerful AI systems.
Stay informed with the latest trends in Web3, blockchain innovation, and cybersecurity updates at 3verseTV
You need to login in order to Like
Leave a comment