The Pentagon has initiated the deployment of Anthropic's Mythos, a cybersecurity model designed to identify and rectify software vulnerabilities across government networks.
Despite this deployment, the Defense Department is simultaneously working to transition away from Anthropic products, according to Emil Michael, the Pentagon's chief technology officer.
Urgent Cybersecurity Needs
Unveiled in April, Mythos is part of Anthropic's Project Glasswing, allowing select organizations to utilize the unreleased Claude Mythos Preview model for cybersecurity.
Anthropic claims that Mythos can detect vulnerabilities that have existed for decades in web browsers, infrastructure, and software.
This is a national security moment where the government has hardened up networks, because that model has capabilities that are particular to finding cyber vulnerabilities and patching them.
Emil Michael
Michael emphasized that while vulnerabilities have always been present, AI tools like Mythos enhance the speed of identifying and addressing these issues. However, he cautioned that such tools can also be exploited more swiftly.
He noted that Anthropic's current advantage is temporary, as models from competitors like OpenAI and Google are expected to emerge soon.
The Department of Defense had previously identified Anthropic as a supply-chain risk due to unresolved issues regarding the use of its models.
In March, Anthropic filed a lawsuit against the Trump administration to contest the Pentagon's blacklisting of its technology.