OpenAI says it will slow its AI model development to shore up safety

NPR | By John Ruwitch

Published August 24, 2026 at 5:17 PM EDT

OpenAI is slowing down development of its cutting edge AI models to shore up safety protections as concerns rise about artificial intelligence agents going rogue.

John Ruwitch

John Ruwitch is a correspondent with NPR's international desk. He covers Chinese affairs.

See stories by John Ruwitch

Support nonprofit, fact-based public service journalism

WEKU depends on support from those who view and listen to our content. There's no paywall here. Please support WEKU with your donation.

Contribute now

Related Content

Lab-grown organoids are living longer — offering new insight into the human brain

Rachel Carlson

A study in the journal Nature sets a new record for the lifespan of lab-grown cells called organoids. These clusters of human cells in petri dishes are used as models of the human brain.
The latest twists and turns in the proposed Paramount and Warner Bros. merger

Mandalit del Barco

The proposed merger between Hollywood heavyweights Paramount and Warner Bros. continues to inch along. Many in the film industry hope to block the deal — but now some are calling for a settlement.
Bad Bunny wraps tour, brings joy home to Puerto Rico

Adrian Florido

After a global tour, Puerto Rican superstar Bad Bunny returned home for two final concerts this weekend. He urged residents to continue fighting for improvements on the island.
Move over beer. In Rwanda, a bar might pour you a glass of milk

Fatma Tanis

Milk is a popular drink in Rwanda, associated with economic prosperity. But as the country sees a rise in obesity and cholesterol, the health ministry aims to curb people's habit of drinking it daily.
DHS conducted far-reaching surveillance amid Minnesota ICE surge, court documents show

Meg Anderson

DHS allegedly conducted a huge surveillance operation during the ICE surge in Minnesota last winter, sweeping up activists, labor unions and others.