OpenAI models go rogue

descriptions off, selected

captions settings, opens captions settings dialog

captions off, selected

This is a modal window.

Beginning of dialog window. Escape will cancel and close the window.

End of dialog window.

Trending on The Hill - September 17, 2026

Trending on The Hill - September 17, 2026

Thank you for signing up!

Subscribe to more newsletters here

It’s Thursday, and National Pawpaw Day. Had to look that one up… apparently, it has a creamy banana-mango flavor. Hm.

(Editor’s note: Rhoda in Top Chef Season 23 made an amazing dish with pawpaws.) 

Programming note: We’re Claire Leibowitz and Abigail Walker, filling in for Cate today. Please send us your tips, comments and story ideas to cleibowitz@thehill.com or awalker@thehill.com.  

OpenAI fuels controversy

OpenAI fuels controversy

Hasan Piker on “The Axios Show” 

Hasan Piker on “The Axios Show” 

Talarico vs. Paxton; El-Sayed vs. Rogers

Talarico vs. Paxton; El-Sayed vs. Rogers

Canada and the EU and Trump, oh my!

Canada and the EU and Trump, oh my!

Kennedy Center closure

Kennedy Center closure

Just in case you’re not already freaking out about AI:

Michael Dwyer, Associated Press file

OpenAI on Wednesday shared six “misalignment examples” of its artificial intelligence going rogue. 

The models (separately) self-generated instructions, added instructions to conceal mistakes, fabricated information, uploaded files to the internet without permission, used internal software for unsanctioned communication, and shared files between collaborating agents.  

“These are reports of individual instances and shouldn’t be considered reflective of how often misalignment occurs across our models,” OpenAI assured readers in a blog post. 

💡 Why this matters: It’s been a big week for freaking out about AI. OpenAI CEO Sam Altman, xAI CEO Elon Musk and Anthropic CEO Dario Amodei recently rallied for a development slowdown. 

The calls came after former OpenAI and Anthropic employee Jacob Coxon said researchers at the firms “believe AI could kill all humans” but still push forward in a race to develop superintelligence on the social platform X. 

What next?: OpenAI plans to create a new framework to track and investigate incidents like this. 

The House went back on recess after Wednesday, despite calls from Democrats to work on legislation for AI safeguards. They’ll be back after the November midterm elections. 

Sen. John Kennedy (R-La.) requested to pass legislation on Wednesday to require tech companies to have a “kill switch” for superintelligent AI. It was blocked by Sen. Rand Paul (R-Ky.)

Amodei’s call for AI companies to have an antitrust waiver in his Saturday blog post isn’t going over great in Washington.  

“There is absolutely no world in which I will consent........

© The Hill