The Trump administration has publicly reversed course on Anthropic, with Commerce Secretary Howard Lutnick declaring on September 1 that the government now trusts the company behind Claude after a bitter dispute that lasted six months and produced a federal lawsuit. This dramatic shift signals how quickly a company's regulatory standing can change, even without resolving the underlying safety disagreements that sparked the original conflict.

What Started the Conflict Between Anthropic and the Pentagon?

The dispute began with Pentagon contract negotiations over how the military could use Anthropic's AI models. The Department of Defense wanted access for "all lawful purposes," including sensitive intelligence work. Anthropic refused to budge on two specific limits: the company would not allow its models to be used in fully autonomous lethal weapons systems, and it would not permit use in domestic mass surveillance.

When those talks collapsed, the fallout escalated rapidly. In February 2026, President Trump directed every federal agency to stop using Anthropic's technology, with a six-month phase-out period. By early March, Defense Secretary Pete Hegseth designated Anthropic a "supply-chain risk," a label that barred defense contractors from using the company's technology entirely. Shortly after that designation, Anthropic took its Fable 5 and Mythos 5 models offline for foreign nationals, citing export control action from Lutnick's office.

How Did Anthropic Fight Back and Win?

Rather than accept the restrictions, Anthropic sued. On August 27, 2026, a federal judge ruled the Pentagon's designation unconstitutional. That legal victory, however, was only part of the story. Winning in court and winning back government trust are two separate outcomes. The court ruling found the Pentagon's action violated the Constitution, but it did not automatically restore any contract or relationship. Lutnick's public statement of trust is what actually signals the government is willing to work with Anthropic going forward, independent of what the court ordered.

Access to Anthropic's models has quietly been restored in stages since the court ruling. Fable 5 access returned to the general public within weeks of the export restriction. Mythos access came back for a smaller group of approved customers. Neither restoration was accompanied by a public announcement at the time, only a private return to normal operations.

What Does This Mean for Businesses Using AI in Regulated Industries?

This reversal offers a real data point on how fast a company's standing with regulators can shift, in either direction. Anthropic went from a formal national security blacklist to a public statement of trust in under six months, without a change in the underlying safety commitments that caused the original dispute. That same fast-moving pattern shows up in Anthropic's own product roadmap, where the company has been releasing new Claude models at a rapid pace.

For businesses operating in regulated industries or considering government contracts involving AI, this case demonstrates that a vendor's regulatory standing today is not a fixed fact to plan around long-term. Consider these key factors when evaluating AI vendor risk:

• Regulatory Standing Shifts Quickly: A company can move from blacklisted to trusted in months, so vendor selection should account for changing political and regulatory conditions rather than assuming current status is permanent.
• Court Wins Don't Guarantee Business Relationships: A favorable legal ruling does not automatically restore contracts or government partnerships; public statements of trust from officials matter separately from judicial outcomes.
• Safety Commitments May Not Change: Anthropic's core safety positions on autonomous weapons and mass surveillance remained unchanged throughout the dispute, yet the government's stance reversed, showing that regulatory relationships involve factors beyond technical commitments.

Businesses navigating uncertainty around AI vendor selection can work through the tradeoffs with AI services and solutions built around real vendor risk assessment, not just feature comparisons or current regulatory status.

What Else Is Happening With Anthropic's Models?

Beyond the political reconciliation, Anthropic has been making significant technical improvements to its Claude models. The company recently made Claude Fable 5.1 and Claude Mythos 5.1 generally available. Fable 5.1 costs about 25 percent less than Fable 5 for typical workloads and up to 45 percent less for agentic tasks, which are AI systems designed to take actions autonomously. The cost reduction comes from cheaper cache-read pricing, a technical optimization that reduces expenses when the model reuses previously processed information.

The new Fable 5.1 model also performs better on technical benchmarks. It scored 52.6 percent on Terminal-Bench-Science, compared to 24.7 percent for the earlier Fable 5 model. On CursorBench, Fable 5.1 achieved 73.4 percent versus 70.5 percent for Fable 5. Additionally, cybersecurity safeguards in the new model produce 60 percent fewer false positives, and the model's writing is less verbose, meaning it uses fewer words to convey the same information.

Anthropic also introduced Enterprise Frontier Safeguards, a program that stores customer data on infrastructure the customer controls instead of Anthropic's cloud servers. This addresses privacy concerns for enterprises handling sensitive information.

It is important to note that Anthropic still faces a separate, unresolved case in the D.C. Circuit under a different statute, and no new government contract has been publicly announced since Lutnick's comments. Businesses watching this relationship should treat the public statement as a signal of direction, not a finished resolution, until an actual new agreement gets announced.