Guardrails

For fifteen years, the Knoedler Gallery sold $80 million in forged paintings that passed expert authentication. The experts assessed appearance, not provenance. The AI security industry is repeating the pattern with guardrails. Prompt injection and hallucinations share the same root cause in the attention mechanism. The fix is not a better filter. It is a provenance system.

Agentic AI treats every tool-call the same way a 19th-century railroad treated every mile of track: one binary safety mechanism regardless of risk. Alert fatigue defeats the human in the loop, and a single hallucination inherits the developer’s full identity. The solution is not to remove the human but to stop asking the same question regardless of the stakes.