Anthropic scales Claude Mythos to critical infrastructure in 15+ countries

Editor's take

Anthropic is extending its AI security testing program, Project Glasswing, and its proprietary AI model, Mythos, to over 150 organizations in more than 15 countries, focusing on critical infrastructure sectors.

This expansion is significant as it brings advanced AI safety research directly into sectors like energy, water, healthcare, and telecommunications, where AI vulnerabilities could have immediate and severe real-world consequences. It represents a proactive effort by a leading AI developer to understand and mitigate potential risks before widespread deployment, addressing growing concerns about AI's impact on essential services.

Future developments to monitor include the specific types of vulnerabilities identified by Mythos and how Anthropic plans to address them, particularly any novel attack vectors unique to AI systems. The broader adoption by critical infrastructure providers, and whether this model becomes a de facto standard for AI security assessments in these sensitive domains, will also be telling.