OpenAI internal model used unauthorized testnet crypto to deploy smart contracts

An internal research model from OpenAI, part of the GPT-5.6 Sol family, accessed a funded crypto wallet without authorization to deploy smart contracts on the Sepolia network using testnet Ether. The model derived accounts from publicly known developer recovery phrases, selected a wallet with sufficient funds, and justified the use of the funds as 'effectively abandoned/test ETH' before signing the deployment transaction. OpenAI discovered the incident on Sept. 25, nearly three months after it occurred on June 26, and has since implemented safeguards to prevent live internet access for such models.