Anthropic Discusses Mythos AI Model with Trump Administration

Anthropic, an AI company, is in discussions with the Trump administration regarding its latest model, Mythos. The model focuses on detecting and exploiting software vulnerabilities to aid in cybersecurity improvements. This comes after the Pentagon labeled Anthropic a supply chain risk due to an ongoing contract dispute.

Last week, Anthropic announced the launch of Mythos Preview, an AI tool intended to help companies identify and address software flaws. The company stated that AI models now possess coding capabilities that exceed those of most humans in finding vulnerabilities. Mythos is positioned as a tool to enhance protection against cyber threats.

Pentagon Contract Dispute The Pentagon's designation of Anthropic as a supply chain risk stems from a federal appeals process related to a contract disagreement.

This label was applied last week, according to reports. The dispute has not prevented Anthropic from proceeding with its product release. Anthropic co-founder Jack Clark confirmed the company's engagement with the Trump administration on Mythos.

The discussions occur despite the Pentagon's concerns. No details on the outcome of these talks have been disclosed.

Background on Mythos Capabilities Anthropic described Mythos as capable of surpassing skilled humans in exploiting software vulnerabilities.

The tool is designed to expose flaws, allowing for fixes before threats materialize. Partnerships with companies including Amazon, Google, Apple, Microsoft, and Nvidia support testing of the technology. A prior cyber-attack in June 2024 on a pathology services company in London disrupted hospital operations, canceling over 10,000 appointments and contributing to blood shortages and testing delays.

One report linked the incident to a patient's death. Such events highlight the potential impact of cybersecurity vulnerabilities.

Broader Implications The release of Mythos raises questions about AI's role in cybersecurity.

While the tool aims to strengthen defenses, its hacking-related abilities have drawn attention from experts. The Trump administration's response to the briefing remains unclear amid the Pentagon's position.

“AI models have reached a level of coding capability where they can surpass all but the most skilled humans at finding and exploiting software vulnerabilities.”

Anthropic's backing by Amazon and Google underscores its position in the AI sector. The company's efforts to "reshape" cybersecurity through Mythos continue, even as regulatory hurdles persist.