The era of open-source artificial intelligence is officially dead. Anthropic has restricted its new "Mythos" cybersecurity model exclusively to vetted corporate and state partners via Project Glasswing. The autonomous model is rapidly uncovering thousands of zero-day vulnerabilities across every major operating system.

By hoarding this capability, Anthropic is executing a brilliant manoeuvre of cognitive enclosure. Above the Law notes the model is arguably the greatest cybersecurity threat in history, having defeated decades of expert review. Restricting public access isolates the existential risk while establishing an immensely profitable enterprise tollbooth.

Chief Information Security Officers must now pay Anthropic to defend their networks against the exact vulnerabilities Anthropic discovered. It is the ultimate subscription model. GovTech rightly identifies this as a systemic shift for global cybersecurity, precipitating a capital-intensive auditing supercycle.

Enterprise boards cannot afford to view this as a software update; it is a fundamental perimeter tax. The state-corporate hoarding of agentic AI models ensures that digital security will soon function strictly as a sovereign rent-extraction mechanism.