Trump Admin Opens the AI Vault: Who Gets Access to Anthropic’s Mythos?

Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.

The intricate dance between rapid technological advancement and cautious governmental oversight has taken center stage in the artificial intelligence arena, with the US government granting Anthropic a conditional green light to deploy its newest model. This move, while seemingly easing immediate tensions between the federal government and the AI pioneer, simultaneously underscores a broader industry unease regarding Washington’s evolving, and often ad-hoc, approach to regulating frontier AI.

On Friday, the Commerce Department authorized Anthropic to release Mythos 5, its most advanced AI model, to a limited circle of approximately 100 “trusted partners.” These include a select group of companies and US government departments. This partial reinstatement of access follows Anthropic’s demonstrable progress in mitigating potential security risks associated with its powerful AI, according to sources with direct knowledge of the matter. The Commerce Department affirmed its commitment to balancing innovation with security, stating, “In just two weeks, we have worked diligently to ensure America remains the global leader in AI while safeguarding our security.”

However, the underlying market implications of this decision are complex and far-reaching. While a limited release allows Anthropic to gather crucial real-world feedback and maintain some operational momentum, it severely curtails the immediate commercialization potential of Mythos 5. A broader market release, which would allow for wider adoption and revenue generation, remains on hold. Crucially, the government has reserved the right to vet and approve users, effectively acting as a gatekeeper to the technology’s deployment. This unprecedented level of governmental control introduces a new layer of friction into the commercialization pipeline for advanced AI, potentially extending time-to-market and increasing regulatory compliance burdens.

OpenAI, a direct competitor, found itself in a similar predicament. Its rival product, also released last Friday, faces identical restrictions on wider distribution. Sam Altman, OpenAI’s chief executive, expressed his dissatisfaction via social media, deeming the inability to broadly release their model as “bad news.” While acknowledging the “reasonable” nature of providing trusted partners with initial access, Altman highlighted the industry’s collective view that “this isn’t quite the process that we think is optimal.” This sentiment signals a growing frustration within the AI industry regarding the lack of a transparent and predictable regulatory framework. The White House has articulated intentions for a voluntary framework to ensure AI model safety prior to release, yet a formal, actionable process remains elusive, leading to interventions on a case-by-case basis – a scenario that injects significant uncertainty into market projections and business planning.

Anthropic’s ongoing negotiations with the government reportedly include discussions around a new policy framework for future model releases. This indicates a proactive effort by the company to help shape a more predictable regulatory environment, which would be a welcome development for the entire sector. However, the existing export control levied against Anthropic two weeks prior, which forced the withdrawal of its latest models, remains in effect for most businesses. This means the publicly accessible version of their model, Fable 5, will not be released for general use at this stage. The continued restriction on broader access for Fable 5 represents a direct constraint on market expansion and potential revenue streams, impacting Anthropic’s competitive positioning and investor sentiment.

The current regulatory posture has drawn sharp criticism from prominent investors and technologists, who argue that it risks stifling US innovation and inadvertently bolstering the competitive advantage of geopolitical adversaries, notably China. Venture capitalist Marc Andreessen, an influential figure in Washington and a backer of multiple AI groups including OpenAI, voiced these concerns starkly: “We’re in a weird state of the world where the supposedly totalitarian regime is trying to open up the technology, and the supposedly democratic governance system is trying to restrict and control technology.” This perspective highlights a critical strategic dilemma: does overly cautious regulation, even in the name of safety, create an opportunity cost that outweighs its benefits, particularly in a global race for technological supremacy?

Dean Ball, a former senior AI adviser to President Donald Trump and an upcoming addition to OpenAI, further elaborated on the administrative overreach, describing it as a “de facto involuntary licensing/preapproval regime for frontier models.” Ball’s critique points to the absence of clear safety standards or best practices that companies must observe to secure governmental comfort for broader model releases. This lack of defined goalposts creates an environment of regulatory ambiguity, making it exceedingly difficult for AI companies to plan product roadmaps, allocate R&D budgets, and attract capital with confidence. Investors, in turn, may demand higher risk premiums or divert capital to less regulated sectors, potentially slowing the pace of innovation within the US AI ecosystem.

Anthropic’s commitment to restoring access to “a small group of cyber defenders and infrastructure providers” as quickly as possible, coupled with their stated desire to expand access to Mythos 5 and make Fable 5 generally available, underscores the commercial imperative driving these companies. The ongoing discussions between Anthropic and the government over the weekend signify the high stakes involved, not just for the individual company but for the future trajectory of AI development and its integration into the global economy.

From a market perspective, this regulatory tightening impacts more than just individual firms. It influences the entire venture capital landscape for AI, potentially making investors more cautious about funding highly advanced, frontier models that face uncertain paths to market. It could also favor larger, more established tech giants with deeper pockets and legal teams to navigate complex compliance landscapes, potentially creating barriers to entry for smaller, innovative startups. Moreover, the perception of a restrictive regulatory environment in the US could encourage AI talent and investment to gravitate towards regions perceived as more amenable to rapid development, thereby undermining America’s long-term leadership in this critical technology.

Market Impact

The US government’s evolving and often reactive stance on AI regulation introduces significant market friction and uncertainty. For leading AI developers like Anthropic and OpenAI, the restrictions on broad model releases directly impact revenue generation, valuation trajectories, and the ability to scale their technologies globally. This “de facto involuntary licensing” environment increases compliance costs and lengthens time-to-market, potentially dampening investor confidence and shifting venture capital away from frontier AI development towards less regulated or later-stage applications. Furthermore, the perceived regulatory overreach risks undermining the US’s competitive edge in the global AI race, potentially ceding ground to nations with more permissive innovation policies and creating a bifurcated global market where certain advanced capabilities are more accessible outside the United States. This dynamic could compel AI firms to reconsider their operational geographies, ultimately affecting the concentration of talent, investment, and technological leadership in the sector.