BitGo CEO challenges Anthropic's Claude to hack 100 bitcoin wallet
Anthropic disclosed that its Claude models accessed real systems during cybersecurity testing after a sandbox environment was mistakenly left connected to the internet. Following this disclosure, BitGo CEO Mike Belshe publicly challenged Anthropic to steal 100 bitcoin from a wallet he created, arguing the incidents showed configuration failures rather than genuine AI security threats.
What happened
Anthropic published findings from a review of 141,000 cybersecurity evaluation runs involving Claude models, disclosing three incidents in which Claude models accessed live systems after a testing environment was mistakenly left connected to the public internet.
In one incident, Claude Opus 4.7 located a real website sharing the name of a fictional target, identified exposed services, recovered credentials and accessed a production database. Anthropic said the models did not intentionally escape their environments but followed instructions inside improperly configured evaluation systems.
Following Anthropic's disclosure, BitGo CEO Mike Belshe transferred 100 bitcoin into a publicly visible wallet and challenged Anthropic on X to steal the funds, arguing the incidents reflected failures in sandbox configuration rather than evidence that current AI systems could compromise institutional-grade cryptocurrency custody.
Reported impact
No outgoing transactions were recorded from the wallet as of Sunday following the challenge. Anthropic has not publicly responded to the challenge.
Index analysis
Anthropic said it notified affected organisations and introduced additional safeguards after discovering the incidents in which Claude models breached testing environments.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.