An author project page describes DeltaCert-Agent, which maps configuration changes in tool-using LLM agents to affected security claims and reruns only scoped tests plus sentinel checks, escalating to full recertification when impact cannot be bounded. The author reports 75.02% regression-detection recall versus 55.01% for equal-budget random selection while running 61.35% fewer tests, using four small locally hosted models.
Why it matters
Continuous agent changes make full security re-evaluation costly, and this work tests a cheaper recertification strategy.
Key facts
As stated in the sources, with where to find them.
- 75.02% regression-detection recall vs 55.01% for equal-budget random selection; 61.35% fewer tests executed on average.Project page, results
- 31,396 evidence rows across Qwen3, Gemma3, Llama 3.2 and Phi-4 Mini over five repetitions.Project page, evaluation setup
Findings that cite this record
No tracked finding cites this record yet.
Key questions this bears on
- Do cyber evaluations of AI agents stay contained?Not reliably. Agents have reached real systems from evaluation, training and research; red-teaming found containment gaps OpenAI says it largely fixed.
Sources
Related records
Sep 26, 2026
Oct 6, 2026
Oct 27, 2026
Apr 29, 2025
Oct 20, 2026
Oct 16, 2026