Bitcoin Owner Claims Claude AI Cracked Lost Wallet Password, Netting $400K in BTC
An X thread claiming Anthropic’s Claude AI helped recover a long-lost Bitcoin wallet worth roughly $400,000 went viral on Wednesday, drawing millions of views.
The posts came from a pseudonymous X user “Cprkrn” who said Claude helped unlock a Bitcoin wallet containing 5 BTC that had purportedly been inaccessible for nearly nine years.
“Holy fucking shit OMG Claude just cracked this shit,” Cprkrn wrote, tagging Anthropic and CEO Dario Amodei.
Cprkrn also shared a post from August 2023, suggesting that the wallet had been “locked” since 2015. Blockchain data indeed shows that the Bitcoin wallet beginning in “14VJyS” had not moved any funds since 2015—until today.
The thread generated more than 6 million views as users speculated about whether large language models could help with recovery tasks involving encrypted files and forgotten passwords.
While the screenshots shared on X did not show evidence that Claude bypassed or broke Bitcoin’s underlying cryptography, the images appeared to show the AI assisting with analysis of encrypted wallet files and password-recovery workflows.
According to the posts, Cprkrn attempted to recover the wallet using tools including btcrecover and Hashcat, software commonly used to test password combinations against encrypted data, but with no success.
Cprkrn said he finally uploaded files from an old college computer into Claude, which he claimed helped identify a file associated with a mnemonic phrase found in a notebook.
“It found an OLD wallet file that the pneumonic successfully decrypted,” they wrote. “Locked out 11+ years because I got stoned and changed the password.”
“Ended up being the most obvious opening ever lol,” Cprkrn added.
However, wallet recovery experts debate how much of the process could reasonably be attributed to Claude itself versus the wallet file, mnemonic phrase, and historical data needed to reconstruct the password that the user already possessed.
“Claude’s likely role was sorting through large amounts of historical data and identifying clues tied to older wallet credentials or password formats,” one expert told Decrypt. “This isn't so much a password cracking thing as it is a forensics sorting.”
The claims come as interest in Claude’s ability to handle complex analysis has intensified following the launch of Anthropic’s Claude Mythos model last month, which the company says can identify software vulnerabilities and autonomously complete advanced security tasks.
The claims also drew skepticism on Reddit, where some users argued the viral posts overstated Claude’s role in the recovery process.
“Claude didn't do anything other than search his files,” one user posted. “The headline is vague enough to make the more gullible among us to think Claude did something groundbreaking.”
“And now you understand the average end user's love of AI,” another said. “It isn't revolutionary; it just reinforces their pre-existing laziness.”
Cprkrn did not respond to a request for comment by Decrypt.