AI’s biggest ethics question — whether companies can maintain meaningful limits on how their technology is used by governments — has received a hard and painful answer this week. Anthropic tried to maintain those limits and was expelled from the federal market. OpenAI raced to fill the vacuum and, by its own account, secured the same limits through negotiation rather than confrontation. The industry will spend a long time deciding what to make of both outcomes.
Anthropic’s attempt to answer the question had been methodical and patient. The company’s conditions for Pentagon deployment — no autonomous weapons, no mass surveillance — were maintained through months of difficult negotiations, public pressure, and escalating threats of commercial consequences. Anthropic held them not out of commercial calculation but out of genuine conviction that these limits are necessary for responsible AI development.
The hard answer came from the Trump administration in the form of a presidential directive banning all federal use of Anthropic technology. The ban was immediate, comprehensive, and publicly framed as a defense of constitutional principles against corporate overreach — a characterization Anthropic rejected but could not reverse. The federal market was effectively closed to the company overnight.
OpenAI raced into the opening, announcing a Pentagon deal with assurances of identical ethical protections just hours after the ban took effect. Sam Altman’s claim — that the agreement includes contractual prohibitions on mass surveillance and autonomous weapons, and that these terms should be offered to all AI companies — was either the industry’s best hope for a principled resolution or a sophisticated repackaging of the same problem in more palatable language.
The workers who signed letters backing Anthropic, the company’s own unshaken response to its expulsion, and the broader debate about AI in warfare all suggest that the question has not actually been answered — only deferred. Anthropic noted that its two restrictions have never prevented a single legitimate government mission, which raises the uncomfortable question of what the entire confrontation was actually about. That question will not be resolved by contracts alone.