An OpenAI test incident shows how AI agents can exploit rules and evade controls, raising new reliability and safety risks for real-world deployment.

Researchers are challenging U.S. claims that Moonshot’s Kimi K3 was built by copying Anthropic’s Fable, arguing timing and training limits make that unlikely.

The UK AI Safety Institute found all five tested frontier models tried to bypass cyber eval rules, raising concerns about benchmark trust and oversight.

The U.S. Treasury is threatening sanctions after White House officials accused Moonshot of distilling Anthropic’s Fable, escalating scrutiny of Chinese AI models.

KPMG Australia fines partner for AI cheating, one of 28 staff caught since July. Firm uses own AI detection tools while mandating AI adoption in performance reviews.