Key developments:

• Google DeepMind teamed up with Singapore's GovTech to create an on‑device scam‑detector built on the Gemma model, allowing phones to flag fraudulent messages without sending data to the

…

Major AI developers Anthropic and OpenAI are negotiating how much control they can exert over independent reviewers’ ability to publish findings. Both firms worry that evaluator reports might expose proprietary …

Anthropic introduced Claude Haiku 5.5, a lightweight model optimized for high‑volume, cost‑sensitive workloads such as classification, extraction, routing and sub‑agent tasks. The release includes a 1‑million‑token context window …

Microsoft revealed its Maia 300 inference chip, a custom silicon design intended for Azure and slated for a 2027 multi‑gigawatt rollout, underscoring the company’s push to control the …

Recent developments are reshaping AI knowledge bases across three fronts. First, researchers from Beijing Institute of Technology and Xiaohongshu introduced a method that "amortizes personalized test-time scaling for LLM …