Signal AI Thursday

🤖 AI News

Google launched Gemini 3.1 Pro, delivering more than double the reasoning performance on the ARC-AGI-2 benchmark while maintaining the same pricing.

Anthropic released Claude Sonnet 4.6 as its new default model, improving coding performance and long-context reasoning — reportedly outperforming its premium Opus model on real-world office tasks.

xAI deployed Grok 4.2 beta with a four-agent architecture that reduces hallucinations by 65% through specialized agents that collaborate and debate conclusions.

Qwen's latest release offers decoding speeds up to 19x faster than prior versions, positioning open-weight systems as serious competitors to closed models.

NVIDIA unveiled its Vera Rubin AI platform at CES 2026, featuring new H300 GPUs engineered for trillion-parameter models.

Scientific breakthrough: MIT researchers developed a generative AI model that streamlines protein-based drug design, potentially saving billions in R&D costs by predicting how synthetic proteins fold with unprecedented accuracy.

🇪🇺 European Section

EU AI Act Implementation Delayed (Again) The European Commission has postponed publication of guidance clarifying high-risk AI systems — marking a second missed deadline. The "Digital Omnibus" package proposes delaying the original August 2026 deadline to between December 2027 and August 2028.

HR AI Compliance Many AI tools used in human resources are classified as "high-risk" under the EU AI Act. Employers must inform employee representatives before deploying such systems.

Copyright Overhaul Proposed The European Parliament proposed substantial changes: GenAI providers would need licenses for copyright-protected works, must disclose an itemized list of training data, and could be barred from EU operation if they fail to comply — even for AI systems trained outside Europe.

Signal AI Thursday — Delivered every Thursday at 7:00 AM UTC