AI Model Restriction
OpenAI Limits New AI Model Release Following Trump Administration's Regulatory Request
OpenAI is restricting its new GPT-5.6 model to just 20 U.S. companies and organizations hand-picked by the federal government, Nexstar Media reported. The White House asked OpenAI to limit the launch after President Trump signed an executive order earlier this month aimed at controlling the release of powerful AI models with advanced hacking capabilities.
OpenAI CEO Sam Altman told staff that the government would be "approving access customer by customer," according to The Information. Altman has also pushed back on the arrangement, saying: "We don't believe this kind of government access process should become the long-term default... It keeps the best tools from users and cyber defenders."
Trump's Executive Order Set the Stage
On June 2, President Trump signed Executive Order 14409, "Promoting Advanced Artificial Intelligence Innovation and Security." The order created a 30-day review window for powerful new AI models before they can be released. It marked the first time a sitting U.S. president had moved to preemptively restrict a commercial AI launch, effectively treating frontier AI like a strategic weapon, according to Axios.
Trump framed the move as a national security necessity. "Advanced AI capabilities make our Nation stronger, but also introduce new national security considerations," he said in a White House fact sheet. Commerce Secretary Howard Lutnick reportedly pushed OpenAI to get multi-agency sign-off before any release, The Wall Street Journal reported.
Anthropic's Hacking-Capable Model Triggered the Crackdown
The regulatory shift was sparked by Anthropic's "Claude Mythos" model. The model was so powerful it was withheld from the public under a program called "Project Glasswing." Pre-release testing showed it could develop working exploits in 72.4% of cases, compared to 0% for earlier AI generations, according to VentureBeat.
The model reportedly found vulnerabilities in every major operating system and browser that had "survived decades of human security review." When a jailbreak was reported in early June, the government moved quickly. It ordered Anthropic to disable global access to Fable 5 and Mythos 5, citing the impossibility of blocking foreign users, Tom's Hardware reported.
GPT-5.6 Launches in Three Tiers, Access Tightly Controlled
On June 26, OpenAI officially launched the GPT-5.6 series under three model names: Sol, Terra, and Luna. Sol is the flagship. Terra sits in the middle tier. Luna is the fastest and cheapest. Sol is priced at $5 per million input tokens and $30 per million output tokens, according to Seeking Alpha. Amazon Bedrock is currently the only confirmed cloud path to access Sol.
Only 20 federally approved partners currently have access to Sol, Forbes reported. OpenAI has not publicly named those partners. Investigators are now asking whether political connections or lobbying influenced who made the list.
Critics Warn U.S. Is Handing China an Advantage
Security analyst Alex Stamos called the earlier Anthropic shutdown a "disaster for the American AI industry," warning it gives China an opening, The Washington Post reported. China recently unveiled GLM-5.2, a competing model that faces far less regulatory resistance at home. A 2026 KPMG survey found only 35% of Americans believe AI's benefits outweigh its risks, compared to 69% in China.
The Cato Institute warned the vetting process is "a step toward a de facto licensing regime" that could choke the innovation Trump promised to protect. Cornell professor Gregory Falco added a different concern: "Government oversight cannot simply mean political review of model outputs," according to CSIS.
Sentiment
NeutralPublishers
21
Articles
20
Reach
21
