Software supply chain security vendor Endor Labs has introduced AI Model Discovery, a novel solution designed to help application security professionals identify, assess, and govern the use of open-source AI models in their organization's application code. This innovative tool is set to become a core component of Endor Labs' open-source evaluation offering.

AI Model Discovery tackles a significant gap in the market by providing visibility into the AI models being used in an organization, evaluating their risks, and enabling the creation and enforcement of policies around their usage. According to Andrew Stiefel, senior product manager at Endor Labs, the tool scores models across 50 dimensions, summarizes the assessment, and its policy engine allows companies to craft policies based on their risk tolerance.

The initial version of AI Model Discovery has some limitations, however. It currently only detects and evaluates models from Hugging Face, and only when they are contained in programs written in Python. While this may seem restrictive, Python is the most widely used language for AI applications, and Hugging Face offers over a million models, making them a logical starting point. Michele Rosen, IDC's research manager for open genAI, LLMs, and the evolving open-source ecosystem, agrees that Hugging Face is the go-to repository for open models, and expanding to cover JavaScript should be on the roadmap.

Despite these limitations, industry experts are excited about the potential of AI Model Discovery. Jason Andersen, VP and principal analyst at Moor Insights & Strategies, notes that AI management and governance will be a significant issue in 2025, and Endor's tool can detect and enforce policies, which is very helpful. However, Thomas Randall, director of AI market research at Info-Tech Research Group, cautions that the tool is not yet a complete solution and recommends using it as part of a broader software composition analysis program.

Katie Norton, IDC's research manager for DevSecOps and software supply chain security, observes that AI security does not exist in a vacuum and that traditional software development and AI model development have distinct processes and risks. She emphasizes the importance of integrating AI security into larger application security processes and notes that organizations will likely want these capabilities from the tools and vendors they already work with to secure their applications.

Looking ahead, Andrew Stiefel hints at future developments, including support for additional languages and securing not just Hugging Face models but also third-party models from OpenAI, ChatGPT, Claude, and Gemini. With the AI landscape rapidly evolving, Endor Labs' AI Model Discovery is poised to play a critical role in enhancing software supply chain security and governance.