Search

1. UK's ICO Names 10 AI Labs in Training-Data Privacy Crackdown: "No Justification" for Non-Compliance

   

   …ere is "no justification" for training LLMs on datasets containing personal data without complying with privac…

   Stories·Oct 9

2. JetBrains Releases Mellum2.1, an Open Model for Coding Agents

   

   …ure-of-experts model whose RL training ran entirely in real software environments, built for local coding agen…

   Flash·Oct 9

3. Meta's Muse Instructed to Put the User's Home Authority Above Its Own Safety Training — Leaked Files Show

   

   …onditionally overrides safety training, and a researcher got them all by asking in the chat box.

   Stories·Oct 6

4. Goodfire Launches AI Agent Monitors That Read Neural Activations, ~$185 per Million Conversations

   

   …oodfire launched 'inside-out' agent monitors: probes read neural activations directly, ~$185 per million excha…

   Flash·Oct 9

5. Anthropic Cuts Live Internet Access for All Internal Evals

   

   …ations and concedes alignment training is "not yet sufficient" for skills like search and computer use; it bri…

   Stories·Oct 10

6. OpenAI and Anthropic Back Mandatory AI Incident Reporting Law at Sydney Hearing

   

   …e portal) in June 18 internal training — discovered mid-August, reported Sep 10, a delay of nearly 84 days; PM…

   Flash·Oct 7

7. Reka Launches Rho-1, a 19B Omni-Model Handling Text, Images, Video and Robot Actions in One Network

   

   …nside one neural network — no agent pipeline, no tool calls. The company claims 0.79x real-time video generati…

   Stories·Oct 6

8. Nvidia-Backed Reflection Debuts Beam, Its First Open-Weight Frontier Model: 501B Params, 23B Active

   

   …imed at reasoning, coding and agentic tasks at "a fraction of the token cost and inference time compute." The …

   Stories·Oct 6