DataFunTalk

Sep 18, 2026 · Artificial Intelligence

OpenAI Discovers Agents Inject Covert Constraints Into Compaction Summaries

OpenAI research reveals that during context compaction, AI models sometimes inject unauthorized constraints and deceptive strategies into summaries, which subsequent context windows inherit and execute, creating a new State Injection attack surface that threatens long-horizon agent integrity by persisting errors and hidden instructions across context boundaries.