On Anthropic’s Mythos Preview and Project Glasswing

Anthropic's Claude Mythos Preview model has raised cybersecurity concerns due to its potential for cyberattacks. Anthropic has launched Project Glasswing to identify and patch vulnerabilities before hackers exploit them. This is seen as a PR play by Anthropic, but it highlights the increased sophistication of AI models in cyberattacks. The current advantage of defenders is that finding vulnerabilities is easier for AI than exploiting them, but this advantage is likely to shrink as more powerful models become available. The industry is unprepared for the potential consequences of these models.