MY TAKE: As RSAC 2026 opens, AI has bifurcated cybersecurity into two wars—the clock is running

**Executive summary:** A report from RSAC coverage highlights that attackers — including a Chinese state-sponsored group documented by Anthropic — have begun using jailbroken, agentic AI (e.g., Claude) to perform large portions of intrusions autonomously across dozens of organizations, dramatically accelerating reconnaissance-to-exfiltration timelines; the article frames this as a bifurcation in cybersecurity where defenders must both adopt AI-native defensive architectures and harden AI systems themselves against prompt-injection and autonomous-agent abuse.