AI Companies Learn the Word No (reason.com)

Anthropic has made a significant decision to restrict access to its powerful AI model, Claude Mythos Preview, due to the identification of "thousands of high-severity vulnerabilities" across major operating systems and web browsers. Instead of a broad public release, the model will be available only to a select consortium of companies, including major tech and infrastructure organizations like Amazon, Microsoft, and Google. This move, part of their Project Glasswing initiative, exemplifies a shift in the AI landscape, where developers are beginning to prioritize safety and ethical responsibility over rapid deployment. By containing the model, Anthropic aims to address critical flaws proactively, rather than risk dissemination to potentially harmful users. The decision also reflects a broader conversation about military applications of AI. Anthropic's CEO emphasized limits on their technology for mass surveillance and autonomous weapons, which sparked a contentious reaction from the Pentagon. As the military insists on "any lawful use," Anthropic is entangled in a legal dispute over being flagged as a risk to national security for their restrictive stance, raising important questions about the balance between innovation, safety, and ethical use in AI development. This evolving landscape suggests that as AI becomes further integrated into military and critical infrastructure, the implications of its deployment must be carefully monitored and managed.