OpenAI's chief executive, Sam Altman, spoke with company staff on Tuesday to address unease following the firm's decision to allow the U.S. Defense Department to use its artificial intelligence systems for classified activities. The pact was made public on Friday, and prompted intense scrutiny from researchers within OpenAI and across Silicon Valley.
The announcement coincided with comments from Defense Secretary Pete Hegseth, who on the same day called Anthropic a supply-chain risk. After critics raised concerns that the original agreement could be interpreted to permit mass surveillance, OpenAI adjusted the terms to explicitly forbid domestic surveillance activities.
According to comments reviewed by a news outlet, Altman told staff he did not regret entering into the Defense Department agreement, while also acknowledging that the timing and manner of the announcement were problematic. He said the rollout felt "opportunistic" and "not united with the field." Those remarks reflected language he used in a memo circulated internally and posted publicly on X on Monday, where he described the deal as appearing "opportunistic and sloppy."
The decision has prompted criticism from AI researchers both inside the company and throughout the broader technology community. Detractors have argued the move amounted to yielding to Pentagon pressure, pointing to contract language that permits the use of OpenAI's tools in "all lawful scenarios."
Key context and company response
OpenAI disclosed the agreement with the Defense Department last Friday. Following immediate criticism, the company modified the agreement to include an explicit prohibition on domestic surveillance. Altman held a staff meeting on Tuesday to explain the rationale behind the decision and to address employee concerns, reiterating his belief in the substantive choice while conceding the execution of the announcement was flawed.
Summary of reactions
- Internal and external AI researchers criticized the deal and its initial language.
- Company leadership responded by amending the agreement to bar domestic surveillance.
- Altman characterized the timing and presentation of the rollout as problematic in both a memo and during the staff meeting.
Note on limitations
The available details focus on the timing of the disclosure, subsequent amendment to the agreement, and expressed staff and industry reactions. Further specifics about the operational terms of the Defense Department's use of OpenAI tools beyond what is described here were not provided in the material on which this article is based.