Topics
All topics →All news
Newest first-
Nadella Calls for an 'Emergency Brake' on AI Models: Assume Compromise, Let Humans Hit Stop
Why it mattersThe Microsoft chief has for the first time publicly framed AI safety around zero trust, demanding a human-controlled emergency brake as standard.
-
Microsoft Launches Decision-1, a Decision Model That Outputs Calibrated Probabilities, Not Text
Why it mattersDecision-1 outputs no text: a single forward pass returns calibrated probabilities for scoring decisions in agent workflows, reportedly 35x faster at P50 and 58% cheaper.
-
UK's ICO Names 10 AI Labs in Training-Data Privacy Crackdown: "No Justification" for Non-Compliance
Why it mattersThe ICO has secured compliance pledges from the 10 biggest foundation-model developers, including Amazon, Anthropic, DeepSeek and OpenAI; the report states that broad interests such as "benefitting humanity" are not a sufficient lawful basis.
-
USA Today Sues OpenAI: $250M+ Damages Sought, "Hundreds of Thousands" of Articles Allegedly Taken for Training
Why it mattersUSA Today Co alleges OpenAI used "hundreds of thousands" of its articles without permission for training and seeks more than $250 million; it is one of "dozens" of copyright suits against tech companies.
-
Google Launches Universal Work Agent Gemini Agent, Running Gemini and Anthropic's Claude
Why it mattersGoogle's flagship work agent launches on both its own Gemini and Anthropic's Claude; "coworker agents" get their own email, calendar and directory identity.
-
Sequoia Leads $60M Series B for Robot-Data Startup Mecka AI
Why it mattersThe $60 million round is led by Sequoia with Nvidia and Microsoft's M12 participating, at a previously reported ~$500 million valuation; Mecka pays people to record everyday tasks to train humanoid robots.
-
Microsoft × NVIDIA Start Selling Surface Laptop Ultra: Runs 120B Models Locally, Top Config ~$7,000
Why it mattersRTX Spark is the first time NVIDIA's Grace Blackwell architecture has reached consumer silicon, claimed to run 120-billion-parameter models locally; the entry S2 has only 24GB of memory, and official pricing is still unannounced.
-
Nous Research Raises $90M Series B at $1.5B Valuation
Why it mattersNous Research is now valued at $1.5 billion with about $158 million raised in total; Hermes has been downloaded about 22.7 million times since February and has overtaken OpenClaw in usage on OpenRouter.
-
Microsoft Ships Its Own Voice Stack: 60-Language Streaming Transcription + Seconds-Level Voice Cloning
Why it mattersAll three transcription and speech models are built in-house by Microsoft; transcription has ~100ms first-word latency at $0.54 per audio hour, and the Flash TTS model runs at 150ms latency for $15 per million characters.
