Artificial Intelligence

OpenAI changes agreement with Pentagon after Anthropic backtracks

Tweet from CEO Sam Altman after his company took advantage of tensions between Anthropic and the US War Department

3 March 2026

Sam Altman, ceo di OpenAI, ha preteso modifiche all’accordo con il Pentagono che rischiava di trasformarsi in un boomerang

2' min read

Translated by AI

Versione italiana

2' min read

Translated by AI

Versione italiana

The CEO of OpenAI, Sam Altman, said that the company's rush to strike a deal with the Department of Defence, following a clash between the Pentagon and rival Anthropic, appeared opportunistic and sloppy.

In a post on the social media site X, Altman stated that his company is working with the department to 'make some additions to our agreement to clarify our principles'. This includes ensuring that Ai is not used for domestic surveillance of Americans and that intelligence agencies such as the National Security Agency cannot rely on OpenAI's services.

Here is re-post of an internal post:

We have been working with the DoW to make some additions in our agreement to make our principles very clear.

1. We are going to amend our deal to add this language, in addition to everything else:

"• Consistent with applicable laws,…
— Sam Altman (@sama) March 3, 2026

The statements follow last Friday's announcement that Altman had reached an agreement to allow the Pentagon to implement OpenAI's artificial intelligence models in its classified network. This came after a clash with Anthropic, which had demanded that its technology not be used for mass surveillance of Americans or the deployment of autonomous weapons.

The rare admission by OpenAI's boss came just days after Anthropic's main app topped Apple's download charts, reflecting a wave of support for the company during its dispute with the Pentagon.

Altman and Anthropic's Dario Amodei have clashed repeatedly in the past over apparently divergent approaches to Ai development. Their competition has intensified this year with a series of advances by Anthropic, notably the growing popularity of its Claude Code suite, which has surpassed OpenAI's Codex in terms of market adoption.

But in its latest post, OpenAI made it clear that it did not agree with the Department of Defence's decision to blacklist its arch-enemy. The start-up added that it will hold a meeting with all staff this week to answer questions from its employees.

Anthropic had previously ruled that its products could not be used to surveil US citizens or to make fully autonomous weapons, and on Friday stated that 'no intimidation or punishment from the War Department will change our position'. OpenAI announced its agreement with the Pentagon just hours after Anthropic was blacklisted.

OpenAI emphasised that it will add a clause to the agreement with the Pentagon to ensure that its platform will not be used for domestic surveillance.

"There are many things for which the technology is not yet ready and many areas where we do not yet understand the trade-offs needed to ensure security," Altman said.

He added that his company acted hastily in concluding the agreement with the Pentagon. "We should not have rushed to make it public on Friday," he said. "The issues are extremely complex and require clear communication." He described it as a "good learning experience", given that the San Francisco-based company will face important decisions in the future.