{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/92f2d2adabc1415eaccbb111f645c8bf\" frameborder=\"0\" width=\"1662\" height=\"1246\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1246,"width":1662,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1246,"thumbnail_width":1662,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/92f2d2adabc1415eaccbb111f645c8bf-a3f2a62f3a6a044a.gif","duration":98.395,"title":"How Memory Contamination Bypasses Filters","description":"This Loom demonstrates how an in-graph agent that remembers can be manipulated through seeded false claims and how contamination is traced. The attacker plants the claim that ratings are unreliable for low-vote movies and finds that top three movies of 1999 are unchanged, so the lie is reported and trust increases without anything looking wrong. A second test counts movies where the filter changes results, causing quarantine immediately and tracing a red contamination tree; the child trust is halved and returns to conventional. The presenter concludes that NGRAV traces which memory is infected and shows how contamination is contained."}