en.Wedoany.com Reported - On April 17, White House Chief of Staff Susie Wiles met with Anthropic CEO Dario Amodei to discuss access to the Mythos model. The White House later stated the meeting was "productive and constructive," discussing opportunities for collaboration and the goal of balancing innovation with safety. Anthropic co-founder Jack Clark confirmed via the X platform on April 14 that the company had already briefed the Trump administration on the Mythos model and expects to continue discussions with government officials regarding future frontier models.
Mythos is a preview version model released by Anthropic through the Project Glasswing program on April 7, positioned above the flagship product Claude Opus 4.6. According to Anthropic's official system card disclosures, Mythos scored 83.1% on the CyberGym vulnerability reproduction test, compared to Opus 4.6's 66.6%; it scored 77.8% on the SWE-bench Pro programming test, a 24.4 percentage point improvement over Opus 4.6's 53.4%; and scored 93.9% on the SWE-bench Verified test, compared to Opus 4.6's 80.8%. In the Firefox 147 vulnerability exploitation test, Mythos succeeded 181 times, whereas Opus 4.6 succeeded only 2 times out of hundreds of attempts. The model's per-task token consumption is approximately 20.4% of Opus 4.6's.
The security capabilities of Mythos have sparked differing actions within the U.S. government. The U.S. Department of Defense designated Anthropic as a supply chain risk company in 2025, restricting its AI models from being used in DoD contract projects. This stemmed from Anthropic's refusal to allow its models to be used for autonomous lethal attacks or large-scale surveillance operations targeting U.S. citizens. The U.S. Department of Commerce's Center for AI Standards and Innovation had already tested the model's offensive and defensive hacking capabilities before Anthropic's official public announcement. U.S. Treasury Chief Information Officer Sam Corcos is seeking access to the model, with the technical team aiming to use Mythos to discover potential vulnerabilities in the financial system. On April 7, U.S. Treasury Secretary Bessent and Federal Reserve Chair Powell convened a meeting with Wall Street executives, urging financial institutions to proactively use the model to detect system vulnerabilities.
The Mythos preview has already discovered thousands of zero-day vulnerabilities, covering all major operating systems and browsers. According to the OpenBSD 7.8 official update log, Mythos discovered a TCP SACK zero-day vulnerability in OpenBSD that had been hidden for 27 years, which had never been found through countless manual audits and automated scans. Mythos autonomously discovered and developed a complete exploit for a 17-year-old remote code execution vulnerability in FreeBSD NFS for less than $1,000. Anthropic provided limited access to Mythos to approximately 48 organizations through Project Glasswing, with core partners including AWS, Apple, Google, Microsoft, NVIDIA, and the Linux Foundation. Anthropic offered participating companies $100 million in usage credits and directly donated $4 million to open-source security organizations.
Prior to this meeting, the Trump administration ordered federal agencies in February to stop using Anthropic products, with Defense Secretary Hegseth labeling the company a "supply chain risk." Anthropic has filed lawsuits in two federal courts. In March, U.S. District Judge Rita Lin of the Northern District of California ruled to partially suspend the enforcement of the designation, but the D.C. Circuit Court of Appeals subsequently briefly upheld it. The legal disputes remain unresolved, and federal agencies and contractors still lack a definitive conclusion on whether they can use Anthropic systems.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
