US intelligence agency tests Anthropic's Mythos for software vulnerabilities
01 May 2026

The US National Security Agency (NSA) is quietly testing Anthropic's latest artificial intelligence system, Mythos, for detecting software vulnerabilities.

The move highlights the growing importance of advanced AI tools in cybersecurity.

The NSA is particularly interested in Mythos's ability to identify weaknesses in widely used software products, including those from Microsoft.

Impressive performance in identifying vulnerabilities
AI capabilities

Insiders told Bloomberg that the NSA is impressed with Mythos's performance.

The system can quickly scan complex codebases and flag possible weaknesses, potentially speeding up a process that usually takes a lot of manual work.

However, it is not clear if this testing has revealed any previously unknown vulnerabilities in the software products tested.

Ongoing comparisons with current tools
Tool comparison

The NSA is comparing Mythos's outputs with its current internal tools and research methods.

This is part of a wider effort to evaluate the effectiveness of advanced AI systems like Mythos in cybersecurity.

However, Anthropic is still cautious about letting this technology spread widely due to fears it could be misused or even weaponized by malicious actors.

Restricted access to Mythos
Access limitations

Anthropic has restricted access to Mythos, giving it only to select organizations.

The company has advised these early users, which include major tech firms and financial institutions, to use the tool defensively. This means identifying weaknesses before they can be exploited by malicious actors.

However, despite its potential benefits, the future of Mythos in government cybersecurity remains uncertain due to ongoing tests and evaluations.

Pentagon flags supply chain risk from Anthropic
Risk assessment

The Pentagon has flagged Anthropic and its technology as a possible supply chain risk.

This comes amid US Defense Secretary Pete Hegseth's earlier indication of moving away from the company's services.

However, Anthropic has challenged this designation in court and recently won a ruling that temporarily blocks restrictions on government use of its technology.

At the crossroads of innovation and caution
Government engagement

Anthropic's CEO Dario Amodei has reportedly discussed with White House officials how the government could safely use versions of Mythos.

As agencies like the NSA continue their tests, the outcome could shape wider adoption of AI-driven cybersecurity tools.

For now, Mythos sits at a crossroads between innovation and caution, offering powerful capabilities while raising equally significant questions about control, access, and unintended consequences.