Sam Altman has made a big bet: that ethical government AI is not just possible but commercially viable. He announced OpenAI’s Pentagon deal with explicit claims of contractual protections against mass surveillance and autonomous weapons, framing the agreement as evidence that AI companies can engage the government without abandoning their principles. The bet was made against the backdrop of Anthropic’s very public fall for attempting the same thing.
Anthropic’s attempt to secure ethical government AI had seemed entirely reasonable: support all lawful military uses of Claude while excluding two specific categories the company considered ethically unacceptable. The company negotiated in good faith for months, made its positions clear and consistent, and offered the government broad and useful AI capabilities within defined ethical limits.
The government decided those limits were unacceptable. President Trump’s ban on all federal use of Anthropic technology, announced on Truth Social in political language designed to cast the company as an ideological adversary, made clear that the administration views AI ethics policies as obstacles to be overcome rather than safeguards to be respected. The fall was swift and consequential.
Altman’s bet in this environment is high-stakes in both directions. If the Pentagon deal delivers on its stated ethical commitments, it will demonstrate that principled government AI is achievable and set a standard for the industry. If those commitments erode under operational pressure — if the contract’s ethical provisions turn out to be negotiable in practice — the bet will have backfired in ways that damage OpenAI’s credibility and the industry’s broader ethical norms.
Hundreds of workers who publicly backed Anthropic before the deal was announced are watching the bet play out with understandable skepticism. Anthropic’s own response — affirming its principles, noting that its restrictions have never harmed a mission, and declining to adjust its position by even a degree — represents the alternative definition of what ethical government AI looks like. That definition involves holding firm; Altman’s involves closing the deal.