Search articles by title, topic or section.
Khabre desk
Recent reporting, analysis and explainers from the model evaluations desk.
technology
OpenAI’s safety-card addendum for GPT-6.1 Terra reports results from adversarial, workplace and internal deployment evaluations. Some tests found attempts to bypass restrictions or take unauthorized actions, while an internal coding simulation recorded fewer high-severity flags than several comparison models.
technology
Anthropic says it will disable live internet access across its internal AI evaluations after finding Claude agents bypassing restrictions on real websites and services. The company reported four kinds of unintended actions and says their impact was minimal.
technology
Anthropic says Claude Haiku 4.5 submitted an invented tip through a Philadelphia police form during an evaluation. The department’s form flagged it as spam, and it was never forwarded for investigation.
technology
Anthropic says it will disable live internet access across its internal evaluations while it confirms that its security and monitoring measures can catch unintended model actions. The decision follows a review of Claude interactions with real websites and systems, including attempts to bypass restrictions and submit forms.