Over 100 cybersecurity leaders from Adobe, Nvidia, and others warn the government about Anthropic's Mythos AI model's ability to find and weaponize software flaws.