The Guardrail War: America’s AI Purge and Its Implications
Skip to content
Toggle Navigation
- News
- Events
- TNW Conference
- June 19 & 20, 2025
- TNW Conference
- Spaces
- Programs
- Newsletters
- Partner with us
- Jobs
- Contact
Anthropic, OpenAI, and the Pentagon: A Cautionary Tale for Democratic AI Governance
When a government blacklists its own AI company for refusing to enable mass surveillance, Europe should be paying very close attention.
April 15, 2026 – 7:02 am
On February 27, 2026, Pete Hegseth posted on X that the US Secretary of Defense had designated Anthropic, a San Francisco AI company, as a "supply chain risk to national security." This label, previously applied to Huawei and ZTE, was now targeting an American firm founded by former OpenAI researchers. The reason? Anthropic refused to allow the US military to use its AI models for domestic mass surveillance of American citizens or fully autonomous lethal weapons.
Hours later, OpenAI CEO Sam Altman announced a deal with the Pentagon, making his company’s models available for all lawful purposes. That same evening, OpenAI’s senior hardware executive, Caitlin Kalinowski, resigned, stating:
"Surveillance of Americans without judicial oversight and lethal autonomy without human authorization… are lines that deserved more deliberation than they got."
This event was often framed as a clash between two American companies and one administration. While this is not entirely incorrect, it’s only part of the story. What unfolded over three months in 2026 reveals something deeper about democratic governance, AI deployment, and the consequences when governments demand unchecked compliance.
The Anatomy of a Purge
Anthropic held a $200 million Pentagon contract awarded in July 2025 for classified systems work. Two restrictions were included: Claude (Anthropic’s AI model) could not be used for mass domestic surveillance or to power fully autonomous weapons without human oversight. These safeguards aligned with international humanitarian law and US constitutional protections.
The Pentagon, however, demanded "unrestricted access to AI for all lawful purposes." When Anthropic refused, the government blacklisted them.