AI News Roundup: Sep 7, 2026

This week:

• Four major AI models suffer rare overlapping downtime
• NYC To Ban AI in Education
• Instagram puts new limits on undisclosed AI profiles

Four major AI models suffer rare overlapping downtime

2026-09-04

Your phone just went dark for a few hours. Not the battery. The four biggest AI models in the world decided to take a nap at the exact same time. ChatGPT, Claude, Grok, and Gemini all hit elevated errors between 9am-12pm.

It feels like the early internet, but it is just a growing pain. When we first wired homes for electricity, the regional grid would brown out while engineers figured out load balancing. I remember when Webex went down in the early days, and (som) customers would smile and call it a vacation day ;) AI for the rest of us means keeping a paper notebook and a backup plan in your drawer.

Sometimes reliance on a new tool (that didn't exist a few years ago) grows faster than the infrastructure to support it.

Read the full story on arstechnica.com

NYC To Ban AI in Education

2026-09-02

New York City banned student-facing generative AI for 600,000 public school kids in grades K-8. The moratorium lasts a year. High schools get limited pilots and literacy modules. Everyone else gets a pause button.

While it's tempting to freeze the status quo, technology does not wait for city hall. Teachers are already using AI to untangle grading piles that grow longer every semester. AI has already entered the classroom. what remains to see if artificial brains can help teachers better develop human ones.

💡 Read my deep dive: Is AI the Future or the End of Education?

Read the full story on nyc.gov

Instagram puts new limits on undisclosed AI profiles

2026-09-01

Less than a week after Meta agreed to an $18 billion settlement over teen safety, Instagram announced it’s renaming its AI labels and limiting the reach of accounts that hide synthetic profiles.

Probably a good thing, now that we are scrolling past wellness influencers and dating promoters who look completely human. The new policy is blunt. If a profile is built by AI, it gets a label. Skip the label and your reach drops. It’s like a grocery store finally putting nutrition labels on every shelf.

I'm glad to see the move back to growing trust now that almost everything looks real. Even though I think I'm pretty savvy, I've been tricked to stop scrolling to watch something that wasn't real. How about you?

Read the full story on techcrunch.com

Want these updates as they happen? Follow me on LinkedIn or X.