OpenAI launches new system the Trump administration initially put on a leash
TL;DR
OpenAI is publicly releasing GPT-5.6 Sol on July 9, 2026, two weeks after the Trump administration initially asked the company to restrict access to government-approved partners. OpenAI says the U.S. government raised no objections to the launch. The model is now open to companies and individuals, not just selected partners. GPT-5.6 Sol is described as especially strong at hacking and security work. That capability is exactly why it is drawing closer political scrutiny.
Nauti's Take
For teams, this is first a security and control test. Anyone evaluating GPT-5.6 Sol should start with bounded red-team scenarios, logging, role permissions, and data-sharing rules before letting the model touch production workflows with code, cloud, or customer access.
Briefingshow
The episode shows how quickly powerful AI models can move into a gray zone between product launch, cybersecurity and government oversight. Officially, there is no formal approval requirement. In practice, it signals that Washington is watching models with hacking potential much more closely.
Sources