📰 News Alibaba Banned Claude Code — and the Reason Is a Story About AI Espionage
Alibaba told staff to stop using Claude Code starting July 10. The reason traces back to Anthropic's own covert user-identification experiment.
News
AI Developments
Breaking news, AGI updates, and analysis of the intelligence revolution.
📰 News 
📰 News Australia's Health Department Just Warned Doctors About AI Scribes — and NZ Should Be Watching
AI scribes are recording, transcribing, and summarising GP visits across Australia. The health department says it's worried. NZ has no equivalent warning.
📰 News GPT-5.5 Is Caught Clustering Its Reasoning at Exactly 516 Tokens — and It's Getting Worse
44% of GPT-5.5's longer reasoning runs end at exactly 516 tokens. In February it was 0.11%. Something changed.
📰 News The NHS Just Put AI Triage in 200,000 Patients' Pockets — and Doctors Are Worried
The NHS is rolling AI triage into its app for 200,000 patients. A Sussex trial cut call queues 29%. But nurses and policy leaders say the productivity claims are overstated and patient privacy remains untested.
📰 News OpenAI Never Visited the Site — The £30 Billion Stargate UK Was a Press Release
The UK government said £30bn was coming to North Tyneside. The Guardian just proved £20bn of it was invented.
📰 News Simon Willison Shipped sqlite-utils 4.0 Mostly Written by Claude Fable — for $149.25
Simon Willison used Claude Fable to ship sqlite-utils 4.0rc2 — 34 commits, 30 files, $149.25 in API costs. A data-loss bug Fable found, two more GPT-5.5 caught. Cross-model review is now standard practice.
📰 News Serious Vulnerability Disclosures Spiked 3.5x After Claude Mythos — the Patching Backlog Is Now the Crisis
Claude Mythos found 10,000+ vulnerabilities under Project Glasswing. The CVE data shows the ecosystem is now disclosing 3.5x more serious flaws than before — and the patching capacity hasn't scaled.
📰 News GLM 5.2 on AMD MI355X Hits 2,626 tok/s — at Half the Cost of Blackwell
AMD's MI355X delivers 80% of NVIDIA B200 inference performance for GLM 5.2 at half the cost. Wafer.ai's MXFP4 quantization and sglang tuning show the CUDA moat is now a support problem, not a software problem.
📰 News NSW Ministers Joked About Skynet While Courting OpenAI — Then Redacted Their Own Enthusiasm
The NSW government wanted OpenAI in Sydney. The tech minister's staff also joked they were headed for a Skynet situation in 5 years. The emails reveal what governments actually think about AI — when they think no one's watching.
📰 News The Right to Run Your Own AI Just Went Mainstream — and the Fight Is at the State Level
Right to Intelligence frames local AI as a civil liberty — download, run, modify, share open models without a platform's permission. The campaign launched at the US state level, and it's hitting a nerve.
📰 News SK Hynix Is About to Launch One of the Largest US Share Sales in History — and the Fee Is Only 0.5%
The world's dominant HBM maker is listing in the US. The 0.5% fee signals a seller's market — banks are competing for the prestige, not the payout. This is what AI chip sovereignty looks like when it goes public.
📰 News Seven US States Are Recalling Officials Over Datacenters — and $130 Billion in Projects Is Now Stalled
The US datacenter backlash has gone electoral. Seven states launched recall efforts in May-June 2026, with bipartisan opposition to AI infrastructure being shoved into communities without disclosure.