Mean Time to Epistemic Failure: The Autoimmune Paradox and 72-Hour Decay Dynamics in Autonomous AI Systems

We introduce Mean Time to Epistemic Failure (MTEF), the first temporal reliability metric for measuring how long autonomous AI systems maintain factual coherence under iterative processing. Through controlled telephone-game experiments across four major language model architectures, we demonstrate that multi-agent AI systems exhibit predictable epistemic decay following quantifiable thermodynamic regimes. Our principal finding—the Autoimmune Paradox—reveals that governed systems decay 2.2x faster than ungoverned systems.