OpenAI Limits New Model Access To Trump Approved Customers

OpenAI and Anthropic have restricted access to certain new artificial intelligence models to customers approved by former President Donald Trump while the companies conduct a cybersecurity review, according to recent reports.

The limits apply to new model offerings from both companies and affect who can use them during the review period. The reports describe the approved users as “Trump-approved customers,” and indicate the restrictions were put in place as the review proceeds.

The development involves two of the most prominent U.S. AI companies. OpenAI is the maker of ChatGPT and related AI products, while Anthropic is a major competitor known for its own AI models. The reports tie the access limits specifically to a cybersecurity review, signaling the companies are treating the release and distribution of these models as a security-sensitive process.

Restricting access to new AI models can shape how quickly the technology reaches businesses, researchers, and developers. New models often bring upgraded capabilities, which can be important for software development, customer service tools, data analysis, and other commercial and public-facing uses.

Because OpenAI and Anthropic supply AI systems that can be integrated across many products and services, changes in availability can ripple beyond the companies themselves. Customers relying on the latest models may face delays, and competitors and partners may need to adjust deployment schedules depending on what access is permitted during the review.

The reports’ description of “Trump-approved customers” places the restrictions in a politically charged frame, raising questions about how approval is determined and how the companies are managing access while the cybersecurity process is underway. The existence of a review also underscores the broader scrutiny around advanced AI systems and how they are protected, distributed, and controlled.

For OpenAI and Anthropic, the move highlights the balance companies are trying to strike between releasing more capable AI tools and limiting exposure during periods of heightened security assessment. Even temporary limits can become a significant operational decision when models are in high demand and widely embedded into other services.

What happens next will depend on the outcome of the cybersecurity review and whether the companies expand access beyond the current approved customer set. Any change in access policy would likely affect product rollouts and customer planning, particularly for organizations waiting to test or deploy new model versions.

For now, the companies’ decision means the newest models remain available only to a narrow slice of users as the cybersecurity review continues, with broader access contingent on whatever steps follow from that process.