Agent Memory Poisoning: Why Content Screening and Provenance Ranking Can't Stop Persistent False Information

Measuring the durability of false information in agent memory systems and why traditional defenses fail against plainly worded assertions.