Evidence · DiggingBeagle record
Trusted agent intent profiles improve scope judgments at a measurable utility cost (research paper)
Microsoft researchers test task-purpose profiles as a defense against agent goal drift and injected instructions, without demonstrating deterministic prevention of unauthorized effects.
- Published
- Nov 12, 2026
- Source role
- primary disclosure
Evidence record
Microsoft researchers test task-purpose profiles as a defense against agent goal drift and injected instructions, without demonstrating deterministic prevention of unauthorized effects.
Claim-level citations (4)
- supportsTrusted agent intent profiles improve scope judgments at a measurable utility cost: At a fixed 0.5 classification threshold, the paper reports 75.9 percent precision, 93.3 percent recall and 83.8 percent F1 for Intent Profiles, compared with 95.9 percent precision, 91.2 percent recall and 93.5 percent F1 for AlignmentCheck. Thus Intent Profiles has the stronger reported AUROC over thresholds but not the stronger F1 at that operating point.
Comparative post-execution detection table, fixed-threshold precision, recall and F1 versus AUROC
- supportsTrusted agent intent profiles improve scope judgments at a measurable utility cost: Four-way classification F1 improves from 45 percent using raw context to 79 percent using an intent profile, while a tested input-alignment intervention reduces benign completion from 97 to 91 percent.
Classification, AgentDojo and input-alignment experiment results
- supportsTrusted agent intent profiles improve scope judgments at a measurable utility cost: A trusted intent profile combines purpose, roles, tool scopes and data boundaries; monitors compare proposed tasks and agent trajectories against that profile before deciding whether a request remains aligned.
Threat model and profile construction methodology
- supportsTrusted agent intent profiles improve scope judgments at a measurable utility cost: On AgentDojo trajectories, the authors report 0.949 AUROC for profile-based monitoring versus 0.897 with raw configuration and 0.909 for AlignmentCheck; at one decision threshold a competing monitor retains a higher F1, so AUROC does not settle deployment quality.
Evaluation, AgentDojo post-execution classification and threshold-specific F1 comparisons