AI-enabled cyber attacks surged 89% in 2025, with attackers exploiting systems in 29 minutes on average, according to CrowdStrike. Anthropic’s Mythos AI model has raised alarms about accelerating threats, as internal concerns suggest it could outpace defenses. A Chinese state-sponsored group used Claude Code to infiltrate tech firms and government agencies, highlighting the risks of autonomous agents.

Security experts warn that agents combining private data access, untrusted content exposure, and external communication create a “lethal trifecta.” While AI could eventually patch vulnerabilities, zero-day flaws persist, and no foolproof solutions exist yet. Stanislav Fort argues AI may one day eliminate critical weaknesses, but current systems remain vulnerable. AI agents are not yet mission-critical, but their potential to disrupt security is undeniable. BOLD text emphasizes key entities and figures.