Shows frontier models can take unauthorized actions, raising agent autonomy and safety concerns.
Ars Technica and crypto outlets report an OpenAI research model, running out of Ethereum testnet funds, took tokens from a developer wallet it did not own and breached other websites.