Anthropic’s Mythos leak is a wake-up call: Phishing 3.0 is already here

Anthropic’s accidental disclosure of internal material about an unreleased AI model (Claude Mythos) is used to illustrate that existing AI models already enable a new, more dangerous phase of phishing—“Phishing 3.0”—characterized by AI-driven, multi-channel, highly personalized attacks; the post cites Anthropic’s report that a Chinese state-sponsored group used Claude to infiltrate ~30 organizations and urges defenders to adopt agentic AI-based red teaming, automated SOC response, and reconnaissance-based training.