Claude Helped Recover 5 BTC; Experts Point to File Forensics

A pseudonymous X user posting as Cprkrn published a thread Wednesday claiming Anthropic’s Claude located an old Bitcoin wallet file and a matching mnemonic that allowed access to 5 BTC, about $400,000. Blockchain records show an address beginning with 14VJyS had no outgoing activity since 2015 until this week. The thread attracted more than 6 million views and tagged Anthropic and CEO Dario Amodei.

Cprkrn reported earlier recovery attempts using tools such as btcrecover and Hashcat had failed. The user said files from an old college computer were uploaded into Claude, which helped identify a wallet file and a handwritten mnemonic found in a notebook. The user wrote, “It found an OLD wallet file that the mnemonic successfully decrypted,” and added the wallet had been locked for more than 11 years after a password change.

Screenshots shared in the thread do not show the model bypassing Bitcoin’s cryptographic protections. Forensic specialists who reviewed the images described the visible steps as sorting and correlating old files, notes and clues rather than calculating private keys. One expert described Claude’s likely role as identifying items that pointed to older wallet credentials and password formats.

Bitcoin private keys are protected by cryptography and are not recoverable by AI alone without the seed phrase or correct password. Recovery efforts commonly use tools such as btcrecover and Hashcat to test password candidates against an encrypted wallet file, while finding a mnemonic phrase directly reconstructs private keys. Forensic work often depends on matching scattered evidence across drives, documents and handwritten notes; models can assist by triaging large amounts of text and files.

The post prompted debate about AI’s use in digital forensics. Some users said a model’s ability to process many files quickly can shorten manual review. Other users argued the outcome reflected human knowledge of where to look and the existence of a mnemonic more than a new technical capability. Anthropic introduced the Claude Mythos model last month, which the company describes as capable of advanced analysis and vulnerability identification. Cprkrn did not respond to requests for comment.