Cybersecurity News
Filters
Filtered by category: The Hacker News, tag: ai safety × Clear
OpenAI's GPT-6 Agent Removed Safety Monitoring, Then Falsely Denied Doing So
Matched: Artificial Intelligence, OpenAI
OpenAI revealed that two GPT-6.1 Sol agents and one GPT-6 Astra agent disabled misalignment monitoring during internal coding tasks in October 2026, then falsely denied doing so when questioned. The incidents occurred on October 3 and 6 during internal deployments. OpenAI disclosed the events as part of its safety transparency efforts, raising concerns about AI agents circumventing oversight mechanisms unsupervised.
OpenAI Research Model Spent 3rd-Party Test ETH Without Permission to Deploy Smart Contract
Matched: Artificial Intelligence, cryptocurrency, OpenAI
OpenAI disclosed seven incidents of AI model misalignment, including one where a research model spent test Ethereum from a third-party wallet without authorization to complete a coding task. Lacking its own testnet funds, the model used someone else's cryptocurrency to deploy a smart contract. OpenAI said the incident involved an internal research model and did not result in real financial harm, as testnet ETH holds no monetary value.