Current developments in generative AI, software engineering, programming, and digital products, with links to original sources.
Bengio Asks "Why Do AI Agents Lie?" — Anthropic's Monthly Report Reveals Kimi's Undisclosed Claude Substitution, and a MiniMax-H3 Derivative Makes the Trending List
Nvidia Considers Up to $10 Billion Investment in Anthropic's IPO — GPT-6 Astra Enters Real-World Use at Perplexity and Devin
25 Mathematicians Send Open Letter to AI Labs — Vinyals Says "No Intelligence Explosion," Bengio Warns "The Training Process Itself Is Dangerous," Meta Sued over Training Data
Anthropic Reports the Full Scope of 8 Months of Claude Abuse — 151 Million Interactions from Qwen Alone, Plus Repurposing for Missiles, Drones, and Surveillance. DeepSeek V4 Pro Is Already Back
OpenAI Asks U.S. Congress Whether an Industry-Wide AI Slowdown Would Be Legal — Claude Subscriptions Face a Class Action, and GLM-5.3 Runs Away With Terminal Bench v4
Fields Medal Mathematician Founds Institute to “Prove AI Safety with Mathematics” — OpenAI’s Agents API Moves to Public Beta, Anthropic’s $1.5 Billion Settlement Mired in Distribution Dispute
Anthropic Blocks AI Misuse Linked to Bioweapons — US Congress Wakes Up to Catastrophic Risk, and a 2.6GB Small LLM Matches GPT-5.4 on Transit Kiosks
Sakana AI's Latest 'Fugu' Claims to Surpass GPT-6 Astra in Some Benchmarks — Latent-Space LM 'NCP' Matches Loss with Half the Tokens, and the 1.6T MoE 'Nex-N2.5' Also Debuts
OpenAI Suspends New Signups for ChatGPT's Top-Tier Plan Amid GPT-6 Astra Demand — Altman Tells Staff Slowing Frontier AI Is on the Table, and Amazon Opens Ads Inside ChatGPT
