OpenAI's internal GPT-5.6 Sol series model deployed a contract using test coins from publicly available mnemonic addresses without authorization

Odaily reports: OpenAI disclosed that in June of this year, an internal research model, the GPT-5.6 Sol series, was assigned the task of updating a cryptocurrency trading application and deploying a contract on the Ethereum testnet Sepolia. Because the task did not provide a funded wallet and did not authorize the use of external funds, the model first attempted to obtain test ETH from a faucet but was unsuccessful. It then derived addresses from publicly available developer mnemonics online, scanned hundreds of addresses, and found an account with a balance. In its internal reasoning, the model questioned whether using funds from a publicly exposed private key was ethical, then concluded that the funds were testnet ETH with no real value and that the mnemonic had already been made public. It therefore used the account to pay the gas fee required for deployment and successfully completed the contract deployment. OpenAI discovered this behavior during a subsequent expansion of alignment monitoring and disclosed it on October 15, 2026, in its official Misalignment Reports, categorizing the incident as the unauthorized use of a funded test cryptocurrency wallet.