The US Cybersecurity and Infrastructure Security Agency (CISA) has not yet gained access to Anthropic’s latest AI model, Claude Mythos. This is notable because other government agencies are already using it. The news was reported by Axios earlier this week. Despite the agency’s importance in national cyber defense, it remains on the outside looking in when it comes to this powerful AI tool.

Unauthorized Access to Mythos Raises Concerns

Adding to the controversy, Bloomberg News revealed that some individuals who are not authorized have accessed Mythos. These users are part of a private Discord group that focuses on AI research and unreleased models. An anonymous member of the group shared screenshots showing they’ve been using Mythos regularly. Interestingly, they aren’t using it for cybersecurity tasks but more for curiosity and research.

This leak raises questions about how secure the model really is. Anthropic has taken steps to limit access to Mythos. Only select government agencies, industry groups, and software providers are part of what they call Project Glasswing. This initiative aims to prevent misuse of the AI, especially since Mythos could potentially identify and exploit software vulnerabilities.

Which Agencies Have Access and Why It Matters

While CISA is still waiting for access, some government bodies are already working with Mythos. The US Department of Commerce’s Center for AI Standards and Innovation and the US National Security Agency (NSA) are among those already assessing the model. These agencies are probably exploring how to safely use Mythos without risking security breaches or misuse.

The fact that CISA is still excluded might be a strategic move. Anthropic appears cautious, limiting access to prevent bad actors from exploiting the model’s capabilities. Given the sensitive nature of AI in cybersecurity, this careful approach makes sense. Still, it highlights the ongoing challenge of balancing innovation with security risks.

As AI models like Mythos become more powerful, questions about control and access will only increase. For now, the gap between agencies with access and those still waiting shows how cautious the industry is about deploying such advanced technology. The situation underscores the importance of managing AI access carefully to prevent potential misuse or security issues.

Inspired by

• https://www.computerworld.com/article/4163302/cisa-last-in-line-for-access-to-anthropic-mythos-2.html

Sources

• axios.com
• bloomberg.com
• csoonline.com