Anthropic launched Claude Fable 5 as its most powerful model yet, specifically touting its biology capabilities. However, users found the model refuses to answer basic high-school-level biology questions, instead handing queries off to the previous flagship model. The contradiction raises questions about overly aggressive safety filters undermining the model's advertised strengths.
As 2025 draws to a close, Google DeepMind has published its annual review, showcasing eight breakthrough research areas in artificial intelligence. This year…