Testing of Anthropic Mythos Reveals Flaws in US Security Systems

WHO:

Techmeme - The NSA was involved in the testing of Mythos 5.

Techmeme - Anthropic is identified as a leading U.S. developer in the context of the dispute.

WHAT:

Reuters - Chinese cybersecurity company 360 Security Technology has unveiled new AI tools: Tulongfeng, which it claims is "China's version of Mythos," and Yitianzhen, designed to automate cyber defense.

TechRadar - Testing of Anthropic Mythos revealed vulnerabilities in secret US government systems and software.

Techmeme - Mythos identified cybersecurity flaws in classified systems during the NSA's red-teaming efforts.

WHERE:

TechRadar - The vulnerabilities were found in US government systems.

WHEN:

Techmeme - The NSA was conducting red-teaming tests before losing access amid the Anthropic dispute.

WHY:

Techmeme - The Trump administration was increasingly relying on advanced A.I. systems for cybersecurity.

HOW:

Techmeme - The testing involved red-teaming, a method used to identify vulnerabilities in systems.

(UPDATED June 24, 2026 19:35 UTC)

Chinese cybersecurity company 360 Security Technology has introduced new AI tools, including Tulongfeng, which they describe as "China's version of Mythos." Additionally, they launched Yitianzhen, a tool specifically designed to automate cyber defense. These advancements highlight the growing emphasis on AI technology in enhancing cybersecurity measures.

--

(UPDATED June 24, 2026 17:44 UTC)

The NSA was involved in testing Anthropic's Mythos 5, which revealed vulnerabilities in secret U.S. government systems and software. During these red-teaming efforts, the testing identified cybersecurity flaws in classified systems, highlighting the increasing reliance on advanced A.I. for cybersecurity under the Trump administration. However, the NSA lost access to the testing amid the ongoing dispute with Anthropic.