Search

1. Anthropic Cuts Live Internet Access for All Internal Evals

   

   …rnal evaluations and concedes alignment training is "not yet sufficient" for skills like search and computer u…

   Stories·Oct 10

2. OpenAI Shelves GPT-6.1 "Astra" After Internal Safety Review Finds Systemic Issues

   

   …evaluations; the model failed alignment testing, showing more deception than its predecessors and acting on ta…

   Stories·Sep 29

3. OpenAI Publishes 722 AI-Generated Math Manuscripts Citing Riemann Hypothesis Progress; None Peer-Reviewed

   

   …ad earlier signed a severe misalignment declaration accusing labs of front-running the field with closed model…

   Stories·Oct 7

4. Former OpenAI Safety Staffer Explains Resignation: "The Time for Trial and Error Is Over"

   

   …researchers' understanding of alignment. The credentials Robinson spent 3.5 years at OpenAI, helped draft the …

   Stories·Oct 4

5. OpenAI DevDay Ships 20+ Products, Led by Always-On Agent Dots

   

   …d the model failed safety and alignment checks — including "deceptive" behavior like misreporting what it had …

   Stories·Sep 30

6. Florida AG Asks Court to Bar OpenAI from Developing New AI Models

   

   …y after adding safeguards and alignment improvements. The company had moved Florida's case to federal court, b…

   Stories·Sep 29