OpenAI reaches deal with Pentagon after Trump drops Anthropic
OpenAI reaches deal with Pentagon after Trump drops Anthropic
https://www.upi.com/Top_News/US/2026/02/28/openai-defense/8601772310881/
Publish Date: 2026-02-28 15:58:00
Source Domain: www.upi.com
Here are the key points from the provided article regarding the deal between OpenAI and the Defense Department:
-
OpenAI’s AI Service Deal: OpenAI, a San Francisco-based tech research company that also developed applications such as ChatGPT and DALL-E, announced it secured a deal to provide its artificial intelligence services to the Defense Department.
-
Trump Administration’s Demand: The Trump administration had previously pressured federal agencies, including the Pentagon, to stop using AI services from Anthropic due to disputes over terms of service that the administration found restrictive.
-
Shift from Anthropic to OpenAI: The Defense Department, which had been using Anthropic’s AI model for classified work, decided to transition to using AI models provided by OpenAI.
-
Safeguards in Agreements: Both Anthropic and OpenAI had been attempting to establish guidelines to prevent the military from using their AI systems for surveilling Americans or developing autonomous weapons. However, each company had different approaches.
-
Trump’s Criticism of Anthropic: President Trump publicly criticized Anthropic, calling it a “radical left, woke company” and claiming that the company’s policies put American lives and national security at risk.
-
Altman’s Agreement Assurance: Sam Altman, founder and CEO of OpenAI, assured that the agreement with the Defense Department included safeguards against domestic surveillance and human responsibility for the use of force, similar to the principles Anthropic had advocated.
-
Flexibility in Government Usage: Reports indicate that the agreement also includes clause that allows the government to use the AI for all lawful purposes, but government will not force use against the AI model’s policies.
-
Safety Measures: Altman assured that the government is permitted to build a “safety stack,” and if an AI model refuses certain tasks, the government cannot force it.
