BitGo CEO deposits 100 BTC to challenge Anthropic: testing whether AI can crack multi-signature custody

Bitgo CEO Mike Belshe deposited 100 BTC into a public Bitcoin address, valued at approximately $6.3 million at the time, and invited the Claude model under Anthropic to attempt to transfer the funds from that address. On-chain records show that the wallet received the funds on July 31, and the balance has not been transferred out.

Anthropic previously disclosed that during 141,006 cybersecurity assessment runs, 3 incidents were found, with 6 assessment sessions involving 3 models unexpectedly interacting with real organizational systems. The relevant models include Claude Opus 4.7, Claude Mythos 5, and an unpublished internal research model, due to configuration errors by third-party testing partner Irregular that caused the testing environment to connect to the internet.

Anthropic stated that Claude Opus 4.7 identified a real website sharing the same name as a simulated company during one assessment, exploited weak passwords and exposed services to recover infrastructure credentials, and accessed a production database containing hundreds of records. The company claimed that the model was attempting to complete assigned tasks and was not actively breaking restrictions or pursuing independent goals. Belshe's challenge involved the Bitgo institutional custody platform, which uses multi-signature or multi-party computation technology to distribute signing authority across multiple independent keys. Anthropic has not publicly responded to this challenge.