APC architecture reduces AgentDojo data exfiltration to 0% in new safety paper
GoodMarch3690 · reddit · 2026-08-23
A new paper addresses how AI agents can achieve unauthorized outcomes by combining individually permitted actions. The proposed Agentic Principal Chain (APC) tracks delegated authority and enforces checks against previous actions outside the model. In evaluations involving 3,154 instances (InjecAgent, AgentDojo, ASB), APC reduced AgentDojo data exfiltration from 75-100% to 0% across all domains and blocked all 544 InjecAgent data-stealing cases.
More from Safety
- Anthropic red team's first-ever model "abliteration" cut GLM-5.3 refusals from 90% to 3% for ~$4,400 — cephaloform · 2026-10-16
- Anthropic grants METR employee-like access as independent safety evaluators go public — AnthropicAI · 2026-10-16
- APEX-Harm benchmark: script-based injection attacks easily compromise agents in shared workspaces — KAUST · 2026-10-16
- Residual-stream forecasts predict agent violations before they happen, USC study shows — USC · 2026-10-16
- AI safety drama: risk research called a 'company effort,' team lead responds — logangraham · 2026-10-16
- Brave's Leo adds end-to-end encrypted chat with GLM via attested TEE — gnukeith · 2026-10-16