The Trump administration has designated frontier artificial intelligence models as subjects of national security review, triggering immediate changes in how major AI companies launch their most advanced systems. OpenAI delayed and staggered the release of its GPT-5.6 model to accommodate government verification, while Anthropic temporarily shut down its Claude Mitos5 service after export control rules banned access by foreign nationals. The moves signal a shift of AI governance from industrial policy to national security, with direct consequences for corporate strategy and global competition.
Pre-release verification becomes mandatory
President Donald Trump signed the executive order "Promoting Advanced Artificial Intelligence Innovation and Security" on the 2nd of last month. The order requires a 30-day voluntary pre-review for frontier AI models, classified benchmarking led by the National Security Agency, and the creation of an AI cybersecurity clearinghouse. The Department of Commerce is empowered to enforce these measures, directly influencing the launch timelines of companies developing the most capable models.
OpenAI's GPT-5.6, released to the public on the 9th, was originally planned for a simultaneous rollout. Instead, the company previewed the model-composed of sub-models Sol, Terra, and Luna-to select institutions on the 26th of last month before the official launch, following government verification. CEO Sam Altman said he coordinated with Commerce Secretary Howard Lutnick, Treasury Secretary Scott Bessent, and National Cyber Director Sean Cairncross. "Many changes were reflected," Altman told local media. He added that the process is acceptable "as long as the process is easy to understand, fair, and fast," and predicted "things will go much more smoothly when developing the next model." This marks a retreat from OpenAI's earlier position that government approval procedures should not become a long-term standard.
Anthropic's service suspension after export controls
Anthropic faced a sharper conflict. Its Claude model had been the only AI authorized for classified work at the U.S. Department of Defense, but the company was designated a supply-chain risk after refusing to allow unrestricted use of its models for mass surveillance and autonomous lethal weapons. When Anthropic launched its top-tier "Claude Mitos5" and consumer-facing "Fable5," the Commerce Department issued export control guidelines barring access by foreign nationals on national security grounds. Anthropic responded by shutting down the service entirely, citing the difficulty of separating overseas users from foreign nationals within the U.S. The shutdown lasted 18 days; Fable5 resumed on the 1st of this month after the controls were lifted.
South Korea's Institute for National Security Strategy (INSS) noted that Anthropic's decision to implement a broader suspension than the government required shows how the actual impact of regulations depends on a company's autonomous judgment. The INSS also pointed to China's Z.ai launching "GLM-5.2" immediately after the suspension, suggesting that a vacuum left by a specific model can create opportunities for rival nations.
Equity stakes and government relations
The administration's influence extends beyond regulation. Multiple foreign media outlets, including the Financial Times, reported that Altman proposed offering a roughly 5% equity stake in OpenAI during initial discussions with the administration. Altman dismissed the reports as containing "many inaccuracies." Anthropic, meanwhile, is actively working to improve relations by recruiting a public-sector executive with over 20 years of experience.
South Korea's strategic recommendations
The INSS report proposed three countermeasures for South Korea. First, redesign the national AI risk assessment framework around cyber capabilities rather than computational volume, moving away from the current "AI Basic Act." Second, establish AI security governance led by the National Intelligence Service. Third, expand the "Sovereign AI" strategy from technological independence to include continuity of access to frontier models. Yang Ji-soo, an INSS research fellow, said, "future AI policy must consider not only what to develop but also how to secure continuity of access to frontier AI." This analysis echoes discussions in AI for Government about how nations are restructuring AI oversight.
Why this matters for Executives and Strategy
The U.S. government's new controls mean that frontier AI launches now require early coordination with regulators, adding a layer of geopolitical risk management to product roadmaps. Companies developing or relying on advanced AI must anticipate export control scenarios, supply-chain designations, and service disruptions that can arise from national security reviews. The Anthropic case shows that a firm's own ethical lines-such as refusing certain military uses-can collide with government demands, leading to abrupt operational freezes. For strategy leaders, the takeaway is clear: integrate government relations and security compliance into the core of AI product planning, not as an afterthought. This shift underscores the need for C-suite leaders to treat geopolitical risk as a first-order concern in AI roadmaps - a theme explored regularly in AI for Executives & Strategy.
Your membership also unlocks:
