Sam Altman promised to amend the "rolling" contract with the Ministry of Defense to ensure protection for citizens from surveillance.

Pennsylvania Pentagon Rejects Anthropic AI and Switches to OpenAI

After the U.S. Department of Defense (DoD) decided to abandon Anthropic’s artificial‑intelligence models, it quickly signed a contract with its competitor—OpenAI. In response, OpenAI CEO Sam Altman admitted that the deal looked “not very pretty” amid a tense political climate.

Brief Timeline
Date | What Happened | Deal | Altman Publicly Supported Former Colleague Dario Amodei, Anthropic Leader
After rejecting Anthropic | The Pentagon began searching for an alternative and hastily signed with OpenAI. | Recently (Friday) Altman called for a review of the agreement’s terms, adding clauses that prohibit using AI to surveil U.S. citizens. | Monday | A new version of the contract was released, including restrictions on deploying AI in military equipment and guarantees against intelligence agencies’ use of the company’s tools (e.g., NSA).

What Changed in the Agreement
1. Clear Restrictions

- Excludes intentional use of AI to monitor U.S. citizens.
- Clarifies that citizen data must not be sourced from third parties without consent.

2. Technical Barriers

- Models may only run on cloud infrastructure; they are unavailable on final military hardware (not for automatic target selection and destruction).

3. Security Guarantees

- The Pentagon received confirmation that OpenAI’s AI tools will not be used by intelligence agencies.

Public and Employee Reaction
- Over the weekend, a sign reading “No Mass Surveillance!” appeared outside the headquarters in San Francisco.
- Some company employees had already expressed dissatisfaction with the initial contract wording.
- Public discontent led some ChatGPT users to switch to Anthropic’s solutions, which since 2021 emphasize safe AI use.

Conclusion
OpenAI now confirms its readiness to collaborate with the Pentagon by implementing technical and legal constraints that ensure safe AI deployment in the military domain. Although the deal was signed hastily, subsequent changes demonstrate the company’s commitment to transparency and meeting public expectations.