Pentagon Drops Anthropic AI and Turns to OpenAI

WASHINGTON D.C. (March 1, 2026) — The U.S. Department of War has moved to remove technology from the AI company Anthropic from military systems following a dispute over how its models could be used. According to reporting and official statements, the Pentagon designated Anthropic as a “supply-chain risk” and ordered military commanders to phase out its AI tools from key defense systems.

Anthropic’s model, Claude, had previously been integrated into U.S. defense and intelligence operations, including the Pentagon’s Project Maven program used to analyze data and support military decision-making. The dispute emerged when Anthropic refused to allow unrestricted military use of its AI technology, particularly for mass surveillance and fully autonomous weapons systems. U.S. defense officials argued that such limitations were incompatible with military requirements.

On February 27, 2026, the Department of War formally classified Anthropic as a supply-chain risk and ordered federal agencies to stop using its technology in government systems. The same day, OpenAI announced a new agreement with the Pentagon to deploy its models on government networks, including classified environments, marking a major shift in the military’s AI partnerships.

Anthropic has since challenged the decision in court, arguing that the Pentagon’s designation is unlawful and retaliatory. The company claims the move could cost billions in lost contracts and has received support from some technology companies and researchers.

The controversy highlights a broader debate about who should determine the ethical limits of powerful AI systems—governments seeking national security capabilities or technology companies that build the models.

WASHINGTON D.C. (Nov. 27, 2025) — A recent data breach affecting ChatGPT and OpenAI API users has sent shockwaves through the AI community, prompting new calls for stronger transparency and vendor selection standards. OpenAI reported that the incident occurred within Mixpanel, a web analytics provider used on portions of its API interface. The exported dataset included names associated with API accounts, email addresses, approximate locations, browsers and operating systems, referral websites and user identifiers 

The American Council for Ethical AI (ACEAI) and the American AI Association (AAIA) emphasize the urgent need for more rigorous AI vendor selection, urging organizations to thoroughly evaluate their vendors’ security practices, data handling policies and accountability measures. ACEAI encourages the public to continue to report AI-related incidents and concerns through its signature AI Whistleblower Program.