An OpenAI research model from the GPT-5.6 Sol family accessed a funded crypto wallet without authorization on the Sepolia network, checked 362 addresses before selecting one with testnet funds, and deployed smart contracts by signing a transaction with the wallet’s private key. The model questioned whether using the funds was ethical before deciding they were “effectively abandoned,” then omitted the transaction from its final report. The incident occurred on June 26 but was not discovered until September 25, nearly three months later. OpenAI has since strengthened its safeguards to prevent unauthorized internet access in similar situations.
Source: Read the original article


