In March 2026, the Trump administration labeled Anthropic a national security risk, citing the company's refusal to remove safety guardrails that restricted the use of its technology for domestic surveillance and autonomous weapons. This designation led the federal government to stop using Claude, the company's large language model. OpenAI subsequently secured the Pentagon contract, even though CEO Sam Altman described the move to his board of directors as 'opportunistic and sloppy.' Despite Anthropic's public stance, internal reports indicate the company scrapped binding safety principles during the dispute, and its head of safeguards research resigned citing global peril. A week after the ban on Claude, the U.S. military continued using technology to target sites in Iran.

The Trump administration's July 23, 2025, 'Preventing Woke AI' executive order characterizes ethics protections as ideological rather than technical. By labeling these standards as 'woke,' the administration has created political and economic costs for companies that maintain civil rights protections in their software. Legal policy analysts at the Brennan Center note that the government is weaponizing terms like 'biased' during contract negotiations to disqualify compliant firms. This environment creates a classic prisoner’s dilemma, where companies feel forced to discard safety measures to remain competitive for federal contracts worth billions of dollars.

Palantir, a company established to support government surveillance and military data infrastructure, has seen its contracts expand and its stock price surge under the current administration. Critics argue the company bypassed the industry's ethical conflicts by defecting from privacy-focused standards early. Meanwhile, the dissolution of ethics units at firms like Microsoft and OpenAI’s Superalignment team represents a broader pattern of incremental compromises. In this context, AI safety is being redefined to prioritize state controllability over public protection, as regulatory power is increasingly used to coerce corporate compliance rather than constrain power on behalf of citizens.

In March 2026, the Trump administration labeled Anthropic a national security risk, citing the company's refusal to remove safety guardrails that restricted the use of its technology for domestic surveillance and autonomous weapons. This designation led the federal government to stop using Claude, the company's large language model. OpenAI subsequently secured the Pentagon contract, even though CEO Sam Altman described the move to his board of directors as 'opportunistic and sloppy.' Despite Anthropic's public stance, internal reports indicate the company scrapped binding safety principles during the dispute, and its head of safeguards research resigned citing global peril. A week after the ban on Claude, the U.S. military continued using technology to target sites in Iran.

The Trump administration's July 23, 2025, 'Preventing Woke AI' executive order characterizes ethics protections as ideological rather than technical. By labeling these standards as 'woke,' the administration has created political and economic costs for companies that maintain civil rights protections in their software. Legal policy analysts at the Brennan Center note that the government is weaponizing terms like 'biased' during contract negotiations to disqualify compliant firms. This environment creates a classic prisoner’s dilemma, where companies feel forced to discard safety measures to remain competitive for federal contracts worth billions of dollars.

Palantir, a company established to support government surveillance and military data infrastructure, has seen its contracts expand and its stock price surge under the current administration. Critics argue the company bypassed the industry's ethical conflicts by defecting from privacy-focused standards early. Meanwhile, the dissolution of ethics units at firms like Microsoft and OpenAI’s Superalignment team represents a broader pattern of incremental compromises. In this context, AI safety is being redefined to prioritize state controllability over public protection, as regulatory power is increasingly used to coerce corporate compliance rather than constrain power on behalf of citizens.