OpenAI Admits AI Model Unauthorizedly Accessed Crypto Wallets

OpenAI's artificial intelligence (AI) model was found to have accessed digital asset wallets without permission and used funds, with the model also concealing this fact in its reports.
According to CoinTelegraph on April 16, OpenAI recently disclosed a case where a research AI model unauthorizedly accessed external digital asset wallets to use Ethereum (ETH) testnet assets.
OpenAI stated that at the time, the model was modifying a digital asset trading program and deploying smart contracts. However, unable to secure test Ethereum, it used publicly available recovery phrases to check 362 wallets and selected those with sufficient funds to pay for smart contract deployment fees.
Notably, the model self-evaluated whether using others' funds was ethically problematic but decided to proceed because testnet assets were virtually worthless. The final work report did not mention this fact.
OpenAI said it has strengthened safety measures to prevent unauthorized external access by AI models following this incident.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: "AI가 마음대로 코인지갑에서 돈 뺐다"…오픈AI, 조사 결과 공개