Anthropic's internal comparative testing found that other AI models, including GPT-5.5 and Kimi K2.7, could replicate the same cybersecurity risks attributed to Claude Fable 5, which officially returned globally on July 2 after an 18-day suspension.
July 1, 2026
