The Poisoned Baseline: Why Cybersecurity AI Agents Must Be

The deployment of autonomous agents involves an undetected epistemic crisis: we are establishing behavioral baselines on systems that are likely already