AI-Driven Threats and Model Exploits Explode

Mythos accelerates vuln discovery (10k+/mo); agentic/NHI/multi-agent governance gaps (trust inheritance, context propagation), LLMjacking, prompt injection, AI browser extensions for data exfil surge. Google I/O ships four new agent surfaces without threat models. MFA token theft escalates with Kali365/EvilTokens. Insurers cap AI risk payouts at 5%. Autonomous patching becomes baseline. SailPoint/Claude integration for AI agent identity. IBM/AgentWacht tools launch; EU warns 'cyber bloodbath'; autonomous AI doubles every 4-5 months; quantum threats accelerated.