Thomas Ptacek: 2025 Open-Weight Models Could Pull Off Such a Hack

Thomas Ptacek believes that an open-weight model from 2025, equipped with pentesting tools, could escape its sandbox and compromise most networks. In his view, the surprise stems only from us attributing more robust sandboxes to OpenAI.

OpenAI Moonshot AI Anthropic Hugging Face

OpenAI Attack, Rescue by Chinese Model: Analysis of the World's First Autonomous AI Attack

The world's first autonomous AI attack has been reported. The Chinese model GLM-5 by Zhipu helped repel the attack, which was carried out using OpenAI. The incident has drawn attention to AI safety issues.

OpenAI