Before the Anthropic-Pentagon deal collapsed over bulk surveillance, it nearly died over two words: "as appropriate." The Pentagon had repeatedly tried to insert vague qualifiers into its contractual pledges, leaving itself escape hatches that would allow the administration to reinterpret the terms at will, The Atlantic has reported.

As reported by "Hvylya", a source familiar with the negotiations told The Atlantic that the Pentagon would pledge not to use Anthropic's AI for mass domestic surveillance or for fully autonomous killing machines, "but then qualify those pledges with loophole-y phrases like 'as appropriate' - suggesting that the terms were subject to change, based on the administration's interpretation of a given situation."

The pattern was consistent enough to alarm Anthropic's negotiators. Each proposed agreement would contain the right commitments on paper - but the fine print effectively gave the Pentagon the latitude to walk them back whenever it chose. For a company whose AI model is the only one currently allowed into the federal government's classified systems, the stakes of ambiguous language were existential.

On Friday morning, Anthropic received word that Hegseth's team would make a major concession by removing the loophole language. The company's negotiators were reportedly relieved. But that relief lasted only hours: on Friday afternoon, Anthropic learned the Pentagon still demanded the right to use its AI for analyzing bulk data collected from Americans - chatbot queries, search histories, GPS-tracked movements, credit-card transactions.

Anthropic's leadership told Hegseth's team that was a bridge too far. The deal collapsed, and Defense Secretary Pete Hegseth swiftly directed the military's contractors, suppliers, and partners to stop doing business with the company. The "as appropriate" language may have been resolved, but the underlying tension - how much latitude the Pentagon would have in deploying AI against its own citizens - proved insurmountable.

As previously reported: OpenAI Rushed to Cut Pentagon Deal Hours After Anthropic Blacklisting, Acemoglu Reveals