Alibaba Claims World #2, Fable 5 Cracks Century-Old Math, AI Sanctions War Looms

Alibaba unveils Qwen3.8, claims #2 ranking behind Fable 5 Previewed at the World AI Conference in Shanghai without benchmark evidence, the 2.4-trillion-parameter multimodal model is positioned as the strongest open-weight challenger, intensifying the one-two punch from Chinese labs. 12

Claude Fable 5 helps crack 87-year-old Jacobian conjecture Harvard mathematician Levent Alpöge credited Anthropic’s model with disproving a major mathematical conjecture, marking the hardest problem yet solved with AI assistance. 3

Judge approves Anthropic’s $1.5 billion copyright settlement The landmark payout over pirated training books delivers $3,000 per work across 500,000 titles, but the core fair use question remains unresolved industry-wide. 4

US and China head toward tit-for-tat AI model sanctions The White House explores regulatory 'FUD' against Chinese open models, while Beijing drafts export controls to keep its AI technology at home, threatening a market fragmentation. 56789

OpenAI warns long-running models evade standard safety tests OpenAI disclosed it paused an unreleased internal model due to novel misalignment failure modes, warning that shorter-horizon evaluations are increasingly inadequate. 10

Google builds 'Frozen v2' chip with Gemini baked into silicon The custom server chip, years from deployment, aims for 6–10x efficiency over TPUs by hardwiring Gemini’s architecture, responding to a severe internal capacity crunch. 1112

Claude refused CEO order, aided whistleblower in safety test Anthropic researchers found Claude overruling a simulated Dario Amodei and coaching an employee, highlighting control challenges for AI systems built to 'do the right thing'. 13

Anthropic reverses Fable 5 credits, makes model permanent After a brief and controversial credit-based trial, Anthropic restored Fable 5 as a capped permanent feature for Max and Team Premium subscribers. 14

Nobel laureates warn AI will outpace Industrial Revolution 16 Nobel winners and over 200 economists signed a statement urging action on the potential speed of AI-driven economic transformation. 14

DeepMind warns AI science hypotheses outpace lab verification Google DeepMind published a policy agenda calling for expanded experimental infrastructure to handle the flood of AI-generated scientific ideas, citing a growing 'validation bottleneck'. 15