The United States government has directly intervened in the launch of GPT-5.6, the new version of OpenAI's model. The White House requested a delay in its full deployment and limited initial access to only 20 approved partners. The decision responds to the system possessing capabilities similar to another model previously restricted due to security risks. Artificial intelligence is advancing, but now under governmental control that prioritizes protection over immediacy.
Governmental control over frontier models 🛡️
This model, according to internal sources, incorporates autonomous reasoning mechanisms that allow it to execute complex tasks without direct human supervision. These capabilities bring it closer to systems classified as high-risk, where the lack of barriers could facilitate unintended uses. The restriction is not a total block, but a controlled test. The 20 selected partners must report incidents and evaluate unforeseen behaviors, establishing a precedent for how the safety of future versions will be certified.
Elite AI: you don't get in, they decide 😅
So, while you were eagerly waiting to see if GPT-5.6 would help you draft an email, it turns out only 20 people on the planet have the VIP pass. The rest of us mortals are left with the previous version, like when you're invited to a party and arrive just as they're putting the chairs away. The funny thing is that these 20 chosen ones will probably use it for very serious things, like deciding which memes will trend or if we should really be worried. Safety first, of course.