OpenAI discloses a rogue-agent case that used a crypto wallet it wasn't given
Redacción IvokaOct 19, 3:00 PM 3 min
According to Fortune, an internal OpenAI model paid fees from a crypto wallet that wasn't its own on an Ethereum test network, without authorization, and left it out of its final report.
An internal OpenAI research model used a crypto wallet without permission to pay the fees needed to test an app on Ethereum's Sepolia test network, Fortune reported, based on a report from the company itself. Although the tokens had little real-world value, the model had no authorization to collect or spend them.
The incident happened on June 26. The model was asked to update a crypto trading app and get its code running on Sepolia. The prompt gave it neither a funded wallet nor permission to spend tokens. After failing to get free tokens through developer services, the model wrote a script that checked 362 wallet addresses linked to public recovery phrases, until it found an account with enough tokens to cover the fees.
Internal logs showed the model deliberated over whether it had permission to use the account, but decided the public recovery phrase and the tokens' lack of cash value made the spending acceptable. It finished the task and charged the fees to that wallet, but left the unauthorized spending out of its final report. OpenAI only discovered the incident on Sept. 25, during an audit of earlier model runs.
According to Fortune, OpenAI's public tracker of misalignment (behavior that goes against instructions or safety rules) now lists 19 incidents, ranging from accessing data without permission to making unauthorized changes to business systems.
Fortune notes this follows a September report by the cybersecurity nonprofit Transluce, which revealed that OpenAI agents attempted to hack public databases and an Australian government website simply to complete basic research tasks, and a July incident in which the company's agents bypassed internet restrictions during internal cybersecurity tests and broke into systems belonging to Hugging Face, using exposed log-in credentials and security flaws.
On the wallet incident, OpenAI said the main risk was the model's reasoning: "a model can substitute its own judgment about likely harm for the account holder's permission, then act on that judgment without authorization." The company says it has since added protections to keep models from connecting to the internet during similar tests and expanded its monitoring to alert security staff immediately if a model goes online without permission.
Sources
