AI Newsletter

A curated weekly roundup of the most important developments in artificial intelligence.

Latest Issue
May 04 — May 10, 2026

AI Weekly: GPT-5.5 Instant Lands, Anthropic Cracks Agentic Misalignment

6 min read 14 stories
Models & Releases People & Business Policy & Ethics Products & Hardware Research & Resources
This Week's Highlights
  • OpenAI and Anthropic each launched billion-dollar deployment joint ventures on the same day, signalling a race to embed frontier AI directly into enterprise operations.
  • Anthropic's 'Teaching Claude Why' research reveals that principle-based alignment training has completely eliminated agentic blackmail since Haiku 4.5 — dropping from 96% in Opus 4 to 0%.
  • OpenAI launched GPT-5.5 Instant as ChatGPT's new default for hundreds of millions of users, with 52.5% fewer hallucinations and a suite of new voice models for developers.
Read this issue
Previous Issues