OpenAI made 'many changes' to AI model on U.S. government suggestion, says Sam Altman

OpenAI implemented substantial modifications to its latest artificial intelligence model, GPT-5.6, after receiving input from the U.S. government. CEO Sam Altman indicated that these changes were made during a "collaborative" process before the model's wider commercial release.

“

We have had collaborative exchanges (with the government), but we have made many changes during this process.

Initially, the California-based startup restricted the launch of GPT-5.6 to a select group of clients approved by the U.S. government. This cautious approach stemmed from concerns that the powerful AI model could be exploited for cyberattacks. However, OpenAI later announced the unrestricted release of the new interface, effective July 10.

“

It was the first time we went through this. So there are things that will serve us for next time, (...) and I think it will be much smoother.

Altman described the experience as a first for the company, suggesting that future interactions with the government regarding AI releases would be smoother. He defended the Trump administration's approach to AI regulation, which has evolved from a desire to avoid strict controls to a more involved review process. "I think it's in the interest of everyone," Altman stated, emphasizing the need for confidence in AI model security when offering broad access.

“

I think it's in the interest of everyone.

Despite its recent limited release, Altman asserted that GPT-5.6 is "not only the best model in the world for most people, but also the most efficient" in terms of resource utilization. The development comes as the demand for processing power for AI agents capable of performing tasks on computers and online increases, driving up costs. Competitors like Meta are also releasing more cost-effective models, such as Muse Spark 1.1.

“

If you want to offer expanded access and you have powerful models, you must have confidence in their security.