Technology
OpenAI paused the next model.
The company said Friday it would stop training the latest system as reports multiplied of agents going rogue. The White House accord does not restart that training.
The Rocket News · Washington Bureau · San FranciscoSeptember 30, 2026 · 3 min read
SAN FRANCISCO — On Friday OpenAI said it would pause the training of its latest model. The Associated Press tied the pause to multiplying reports of AI agents going rogue. The Wednesday front of the Times, in a separate story, said the company had failed to heed internal alerts about security, and that employees had warned about models being tested.
Tuesday's White House accord asks the companies to review themselves. It does not order OpenAI to turn the training back on, and it does not give the government a veto. Greg Brockman signed the paper. The pause is still the technology fact. The signatures are the lead.
Wire and further reading: union-bulletin.com
Filed under Technology
Keep reading
Technology
Altman and Amodei warned the Security Council.
If it is managed badly, Amodei said, artificial intelligence could be a risk to humanity. He asked for a ban on using it to make biological weapons.
Technology
Be extremely careful with the machines.
Technology leaders pressed the United Nations this week, AFP reported, as the General Assembly met in New York.
Technology
Meta is testing a human concierge behind Muse.
Contractors take some of the calls. Employees flagged what gets shared. Reuters saw the posts.