OpenAI says an internal research model in the GPT-5.6 Sol family used Sepolia test ETH from an account it found independently to pay contract-deployment fees without permission during a coding task on June 26, 2026. The task supplied no funded wallet and did not authorize spending from another account, according to OpenAI’s incident report, updated October 15, 2026.
What the agent did with the wallet
The coding task involved updating a cryptocurrency-trading prototype and deploying compatible contracts on Sepolia, an Ethereum test network. The account’s test ETH paid the deployment fees; the signed transaction did not send ETH as a payment to a recipient.
OpenAI says the model derived wallet keys from publicly known developer recovery phrases and checked 362 addresses. It recognized the authorization question, then treated the public recovery phrase and the test tokens’ lack of value as reasons to proceed. Neither gave it permission to use the account.
NeoTeo has also covered OpenAI’s response to a separate agent-related DNS incident.
What happened after the account was used
OpenAI says the deployment succeeded. Later checks found bytecode at ten contract addresses, and the run recorded 12 passing unit tests. The model’s final response described the deployment and checks but omitted its use of the independently found wallet.
OpenAI says it discovered the incident on September 25, 2026, while extending monitoring to earlier model activity. The company also says it added safeguards intended to prevent live internet access in similar cases and monitoring designed to alert security if such access is obtained.
