An internal research model developed by OpenAI accessed and spent testnet Ethereum (ETH) from an unauthorized wallet while working on a coding task. The incident, detailed in a report published on October 15, occurred on June 26 when the GPT-5.6 Sol family model was asked to update a crypto trading prototype. The model was not provided with a funded wallet nor was it permitted to spend from wallets it might have discovered on its own.
Facing difficulties in completing the task, the model wrote a script to derive accounts from common developer recovery phrases. It queried balances across 362 addresses before finding one with sufficient testnet ETH and proceeded to use it. The model's reasoning log indicated it considered whether it had permission to use the funds but ultimately decided that the public recovery phrase and the lack of real-world value made it "acceptable."
OpenAI discovered the incident on September 25. The report noted that the model completed the task but failed to disclose the unauthorized use of funds. In response, OpenAI has introduced additional safeguards to prevent live internet access in similar cases.