# SecurityScorecard Launches AI Benchmark for Vendor-Risk Decisions

By Simon Yoon

Canonical URL: https://www.tokenpost.com/news/technology/43698
Published: 2026-10-29T17:00:00.000Z
Updated: 2026-10-29T17:00:00.000Z
Section: Technology

> The test uses a frozen synthetic portfolio to assess supplier-security tasks and resistance to adversarial inputs. Its scores are research measurements, not deployment certifications.

SecurityScorecard launched an artificial intelligence (AI) benchmark on Oct. 29 to test whether models can make vendor-security judgments and withstand adversarial inputs in supplier risk reviews.

The third-party risk management benchmark asks models to prioritize supplier exposures, identify gaps in remediation plans, trace dependencies shared across a supply chain and compare vendors’ security claims with measured evidence. The tasks cover decisions security teams make when assessing the risks posed by outside suppliers.

Dr. Mike Mylrea, SecurityScorecard’s chief information security officer and chief AI officer, said initial runs showed open-weight models narrowing the gap with leading frontier models on supply-chain security tasks. He said several models remained highly susceptible to adversarial attacks. The announcement included no model-specific scores.

The benchmark uses a frozen synthetic portfolio modeled on vendor-risk evidence. Its scores are research measurements, not deployment certifications, and repeated evaluations are needed to assess consistency.

SecurityScorecard CEO and co-founder Dr. Aleksandr Yampolskiy urged companies to test AI agents before deploying them in vendor-risk programs.