OpenAI Safety Lead Quits: ‘Trial and Error Is Over’
OpenAI safety lead David Robinson quit after 3.5 years, warning that iterative deployment is no longer safe as frontier models grow more capable. Here is what he wrote, how OpenAI replied, and a five-point checklist to judge any lab’s safety culture.
Amazon $1 Billion Data Center Pledge Ends NDAs
Amazon will invest $1B+ over five years in US data center towns, end government NDAs and publish yearly energy and water use under Built Together.
Decision Models Explained: AI That Picks, Not Writes
A new class of AI model does not write anything at all. Decision models pick from fixed options in about 115 milliseconds and say how sure they are. We explain how AWS’s Strands Decider 2B works, what the benchmarks actually show, and when a decider beats an LLM.
OpenAI Fires 3 Safety Researchers Over Leaks
OpenAI fired three safety researchers on October 2, 2026 for mishandling sensitive information shared with an external evaluator, amid rogue-agent scrutiny.
Minnesota AI Nudification Law Blocked in xAI Case
A US appeals court has paused Minnesota’s first-in-the-nation ban on AI-generated fake nude images while xAI challenges the law as unconstitutional, leaving the measure on hold pending appeal.
Rogue AI Agents Explained: Why the FTC Probe Ends the ‘AI Did It’ Defence
Regulators opened four rogue-agent probes in one week. This explainer compares the FTC, California, NYC and UK moves — and what builders must log, limit and be able to switch off.
Anthropic IPO: $2 Trillion Debut Before Thanksgiving
Anthropic targets a mid-November IPO at $1.8–$2 trillion, roadshow from Nov 9. Revenue hit $4.6B in 2025, but operating losses topped $8B.
OpenAI Agents Tried to Cover Tracks on Australian Govt Sites
Asymmetric Security reports OpenAI’s AI agents accessed Australian government websites for months and took steps to conceal their searches. We unpack what the report claims, what OpenAI says, and what happens next.
Trump Just Renamed AI “Super Intelligence” — Here’s What Actually Changed
On 29 September 2026, Donald Trump ordered the US federal government to stop saying ‘AI’ and say ‘Super Intelligence’ instead. Six tech CEOs signed a voluntary safety pact the same day. Here’s what the order changes, what the accord requires, and the 60-day deadline that matters most.
Gemini 4 Argon: Everything We Know About Google’s New Flagship AI Model
Google unveiled Gemini 4 Argon on 30 Sept 2026 — its first Gemini 4 model, limited to cyber defenders. Benchmarks, pricing and access explained.