OpenAI said it will publicly release its highly advanced GPT-5.6 artificial intelligence (AI) model on Thursday, ending a two-week standoff with the Trump administration over national security and regulatory oversight.
Wide release of the flagship model, named Sol, alongside its lower-tier counterparts, Terra and Luna, comes after OpenAI initially bowed to White House pressure last month to restrict access to a small group of government-approved partners. The delay was prompted by federal concerns regarding potential cybersecurity, biological, and coding risks inherent to the advanced technology.
Described by OpenAI as its “strongest model yet,” GPT-5.6 Sol will become widely available to the public, enterprises, and global developers starting Thursday.
“Happy building,” OpenAI CEO Sam Altman wrote in a brief post on social media platform X late Tuesday, signaling the conclusion of additional safety testing conducted by the Department of Commerce’s Center for AI Standards and Innovation.
Friction between Silicon Valley and Washington highlights an increasingly turbulent relationship as the government attempts to navigate AI regulation in real time.
Last month, President Trump signed a sweeping AI executive order establishing a voluntary framework that allows federal agencies to test cutting-edge models for up to 30 days prior to their public debut.
While the administration has maintained that these evaluations are strictly voluntary, critics and industry insiders argue the White House is enforcing an “ad hoc” pre-clearance system. OpenAI openly chafed at the initial deployment restrictions last month, warning that a staggered rollout “keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them.” The company added that such government intervention should not become the long-term default.
Compounding the industry’s anxiety, a White House official actively disputed the notion that the administration had “cleared” or “green-lit” OpenAI’s launch, telling reporters that no such federal permission is legally required. The official stressed that under the June 2 executive order, mandatory licensing is explicitly barred, and the timing of any release “rests entirely with the companies.”
Yet, OpenAI is not the only tech giant to face off with the administration.
Rival Anthropic was recently forced to abruptly pull its advanced Mythos and Fable models from the market following a sharp export control directive from the Commerce Department that banned foreign access. Though the restrictions on Anthropic were lifted last week, the aggressive federal interventions have sparked widespread confusion and backlash among AI policy advocates, who warn that unpredictable government overreach could stifle American innovation.
“This isn’t really about OpenAI clearing a hurdle — it’s the U.S. government establishing a repeatable pre-release review gate for frontier models, the same pattern we just saw with Anthropic’s Claude suspension and reinstatement,” said Shourabh Rawat, vice president of technology at Symphony AI. “For enterprise IT leaders, the practical implication is to stop treating GA dates as firm and start treating regulatory clearance as a recurring vendor-risk variable — right alongside cost and data residency — especially as that review friction gives unrestricted, cost-effective alternatives like China’s GLM models room to gain enterprise mindshare.”
On Wednesday, Anthropic extended its promotional access window for its flagship model, Claude Fable 5, pushing the expiration date to July 12. According to an update on the company’s official support page, the extension allows eligible subscribers to continue leveraging the model’s advanced capabilities as part of their existing tiers. During this promotional window, users can allocate up to 50% of their weekly subscription limits to Claude Fable 5 at no additional cost. Once this threshold is reached, users retain the flexibility to either burn usage credits to maintain access or switch to alternative Claude models to preserve their remaining plan limits.
The extension comes at a critical juncture for the generative AI sector, widely seen by industry analysts as a strategic play to lock down developer mindshare. By lowering the barrier to entry for its premier model, Anthropic aims to anchor its user base ahead of major shifts in the ecosystem, especially as GPT-5.6 comes to market Thursday.
“Enterprise IT just learned that model availability is a regulatory variable, not a vendor roadmap variable. Anthropic’s most advanced models disappeared from the market for three weeks in June – not because of an outage, not because of a pricing change, because of an export control directive,” said Lewis Carhart, CEO of CompAI. “If your AI architecture assumes the model you deployed today is available tomorrow, that assumption is now demonstrably false. Multi-model resilience just went from nice-to-have to a board-level risk item. The flip side is the opportunity: A model that cleared government security testing is a model your auditors and your board sign off on faster. Government review is quietly becoming a procurement asset. The CIOs who win here are the ones who treat ‘regulatory posture of the model itself’ as a line item in vendor risk assessments – most risk teams are still only looking at the provider’s SOC 2 attestation.”
