APREX · EST. 2026
Anthropic's foundational premise is that the company is building some of the most dangerous technology in history — and that this is precisely why it should be the one responsible for developing it. That logic gives Anthropic an unusual degree of latitude: it can act aggressively in the marketplace while simultaneously challenging government authorities, all under the umbrella of accountability.
Commerce Department ordered global model shutdown
In June 2026, the U.S. Department of Commerce issued an export control directive requiring Anthropic to suspend access to its two most advanced models — Fable 5 and Mythos 5 — for all foreign users. That included the company's own employees based outside the United States. Anthropic chose to shut down the models globally in order to ensure compliance.
The government's rationale was that a "jailbreak" method had been discovered that could potentially circumvent the safety mechanisms in Fable 5 and enable the identification of software vulnerabilities. Anthropic disputed the assessment. According to the company, the jailbreak in question produced only minor findings that were already accessible through other publicly available AI models, according to Stratechery.
The incident has been described as one of the most striking examples to date of direct government intervention in the AI race.
Anthropic is refusing to let the Pentagon use AI for domestic surveillance — and is accepting the consequences.
On a collision course with the Pentagon
Running in parallel is an open conflict with the Trump administration and the Department of Defense. Anthropic has declined to allow the U.S. military to use its models for domestic surveillance and fully autonomous weapons systems. The price is a place on a vendor blacklist set to take effect later in 2026.
Pentagon Chief Information Officer Kirsten Davies publicly attacked Anthropic's stance, emphasizing that national interests outweigh commercial considerations. Trump's AI adviser David Sacks has characterized the company as an "AI doomsday prophet" engaged in sophisticated regulatory manipulation driven by fearmongering, according to Stratechery.
For his part, Anthropic CEO Dario Amodei has openly criticized Trump's approach to export restrictions and urged voters to support Kamala Harris ahead of the 2024 presidential election.
Lobbying for millions
Anthropic's safety orientation is not merely an ethical stance — it is also a political strategy backed by substantial resources.
The company has invested heavily in shaping U.S. AI policy since it entered the federal lobbying arena in 2023. In the third quarter of 2025, Anthropic spent more than one million dollars on in-house lobbying, according to Stratechery. The company has backed transparency legislation in several states, including California's SB 53, New York's RAISE Act, and Illinois SB 315.
Safety as competitive advantage
The Stratechery analysis points to a central paradox: Anthropic's self-imposed safety mandate gives the company the political credibility to challenge governments and the military — while simultaneously legitimizing aggressive commercial expansion. The logic is that no one is better positioned to manage dangerous AI than those who take the danger seriously.
Whether this strategy will hold once the blacklisting takes effect and export controls tighten remains to be seen. But for now, Anthropic has chosen to fight these battles rather than compromise with the U.S. defense establishment.