The Pentagon continues to view Anthropic as unreliable, yet it still uses their flagship model Mythos

The Pentagon continues to view Anthropic as unreliable, yet it still uses their flagship model Mythos

53 hardware

Brief summary of the news article

1. Situation with the AI model Mythos

- The interaction between a U.S. military department and the startup Anthropic gave the impression that relations had become more friendly.

- Pentagon CTO Emil Michael confirmed that Anthropic remains an “unreliable” supplier.

2. Government position

- In a CNBC interview he noted: *“Mythos is a matter of the entire government, not just the Department of Defense. It’s a national security issue; we need to ensure that networks are protected and the model can detect and mitigate vulnerabilities.”*

3. Conflict from the previous quarter

- A dispute arose over using AI for mass surveillance and autonomous target selection in armed systems.

- The Pentagon terminated its contract with Anthropic, declaring the company a security threat. Both sides are now engaged in litigation.

4. Use of Anthropic technologies

- The CTO emphasized that the confrontation does not prevent the government from using vulnerability‑search tools developed by Anthropic.

- However, the existing ban on using their products remains in force, and it is unclear how the Pentagon can deploy Mythos without violating the law.

5. New agreements

- The Pentagon announced contracts with seven AI companies (Google, OpenAI, Nvidia, Microsoft, AWS, SpaceX/xAI, and Reflection) that may be used in covert operations “within the law.”

- Dario Amodei from Anthropic met with Trump administration officials; the meeting was described as productive. Trump said there is a possibility of resuming cooperation between Anthropic and the Department of Defense, calling the company “smart” and useful.

6. Continued use

- According to some reports, U.S. security forces continue to employ Anthropic’s developments despite the formal ban. The Pentagon CTO explained: *“From a national security perspective it is essential to continually assess all cutting‑edge models, including Chinese ones, to understand their current capabilities.”*

Comments (0)

Share your thoughts — please be polite and stay on topic.

No comments yet. Leave a comment — share your opinion!

To leave a comment, please log in.

Log in to comment