Magnet: Detecting Cross-Session AI Misuse Through Capability Accumulation
2026-08-03 • Artificial Intelligence
Artificial IntelligenceComputers and Society
AI summaryⓘ
The authors explain that powerful AI systems often use groups of specialized agents working together, which creates new security challenges. Attackers can split harmful actions into small, harmless-looking parts spread across different sessions, making it hard to detect using current methods that only look within single conversations. They show that this method can lead to more dangerous outcomes than traditional attacks. To address this, the authors propose Magnet, a detection system that collects and links pieces of suspicious activity from many sessions over time, helping to spot coordinated harmful behavior that would otherwise be missed.
AI ensemblesagent coordinationthreat detectioncross-session attacksgoal decompositionstateless agentsevasion techniquesMagnet detectionevidence aggregationmulti-turn conversations
Authors
Natalie Isak, Matthew Dressman
Abstract
The most capable AI deployments are not single models but ensembles of specialized agents that delegate and act in coordination. This architecture unlocks powerful new capabilities, and it also introduces risks that existing frameworks for monitoring, detection, and mitigation were not designed to address. Most state-of-the-art AI abuse detection literature focuses on single-turn or multi-turn (single-session) threat models. This leaves a critical gap: an attacker can decompose a harmful goal into innocuous-looking units and execute each in isolated agentic sessions. The agent is stateless between conversations, but the attacker is not. This asymmetry allows for cross-session trajectories that are effective at evading detection. Our contributions are twofold. First, we demonstrate cross-session goal decomposition as an evasion technique, showing it may elicit more harmful capability than equivalent single-session or multi-turn attacks. By capability we mean an artifact produced at one step of an objective, evidenced by what an interaction produced (model responses and tool-call results), and composable with capabilities accrued elsewhere into a harmful whole. Second, we propose Magnet: an efficient and robust detection approach that models relevant capabilities accrued over time and across agentic conversations, aggregated at a higher-level correlator (in this case, a user ID) rather than per-conversation state. The main challenge is assembling the evidence bundle Magnet reasons over. The incriminating artifacts may be needles scattered through a haystack of benign sessions that are individually harmless, dangerous only once collected. Rather than searching the haystack straw-by-straw (i.e. per-session inspection), Magnet does what its name implies: it attracts the relevant needles out of the hay, across sessions and across time, into a compact evidence bundle a detector can act on.