The Week the Most Powerful AI Needed a Government Permission Slip
Week of June 22 to 29, 2026 | Issue 12 | For smart professionals catching up on AI
This week
OpenAI built its most powerful AI yet, and almost nobody is allowed to use it. GPT-5.6 launched on Friday but only about 20 government-approved companies can touch it, the first time Washington has stood between a US AI lab and its own customers. Google lost two of its biggest AI names to rivals in a single week and its stock dropped 7%. Meanwhile, away from the power struggle, an AI earned FDA clearance to spot hidden heart disease from a routine test, summer travel scams got harder to detect, and a $4,500-a-week AI summer camp opened in the Hamptons. The frontier is getting gatekept. The everyday stuff is getting real. Here's what touches your week.
📡 Industry pulse — 60 seconds
The AI industry headlines from this week. One sentence each, so you have something to say when they come up at work.
- OpenAI launched GPT-5.6 (Sol, Terra, and Luna) on June 26, but limited it to about 20 government-approved partners at the Trump administration's request, the first staggered, government-gated release of a major US model.
- Google is bleeding top AI talent. Nobel laureate John Jumper left for Anthropic and star researcher Noam Shazeer went to OpenAI, and Alphabet's stock fell about 7% on the news.
- A report claimed Anthropic's banned Mythos model cracked NSA classified systems in hours, which, if true, explains why the government pulled it offline two weeks ago.
- Oracle disclosed in an SEC filing on June 22 that it cut about 21,000 jobs in a year, explicitly naming AI adoption as a cause and saying the cuts will continue.
- Microsoft and Anthropic both launched "Cowork" tools, AI that does tasks alongside you in your actual files rather than just chatting in a side window.