US demands unhackable AI models: Can Anthropic deliver?

The US government is reportedly pressing Anthropic to deliver something it may not be capable of: unhackable large language models (LLMs). According to insiders, officials from the Department of Commerce, the CIA, and science advisor Michael Kratsios are in talks with the company after it launched Fable 5 without awaiting federal approval. One administration official described the situation as "They screwed us," reflecting frustration over the perceived oversight lapse.

Unprecedented expectations

The demand for "unhackable" AI models goes beyond standard security practices. Current AI systems, including those developed by Anthropic, are designed with robust safeguards but remain vulnerable to sophisticated attacks. The government’s push for absolute security raises questions about whether such a standard is achievable—or even realistic—given the rapidly evolving nature of cyber threats.

Industry implications

This development highlights growing tensions between innovation and regulation in the AI sector. Companies like Anthropic operate in a fast-moving field where time-to-market often conflicts with government oversight. If the US government insists on stricter controls, it could set a precedent for how AI models are developed and deployed globally, potentially slowing progress while attempting to enhance security.

---

Source: The Decoder. AI-assisted editorial synthesis — TechnoExpress.