Anthropic will reportedly release its new AI model “Mythos” tomorrow after alleged delays over cybersecurity risks.

Anthropic's reported release of Mythos shows that AI safety evaluations are struggle-testing actual dual-use risks like autonomous hacking.

* The delay underscores that LLMs are now capable of automated vulnerability discovery and exploit generation, representing a major cybersecurity inflection point.

* Defensive security teams will gain powerful automated scanning tools, but the dual-use nature of these capabilities poses immediate threats if misused.

* The launch signals a market shift where model providers compete on autonomous agentic execution rather than static conversational intelligence.