Open weight AI models are starting to land in production environments the same way that open source software did decades ago - fast, broadly, and largely untracked. Organizations are observing internal teams including open models (from Hugging Face, ModelScope or similar repositories), leaving operational and security/compliance teams with limited visibility into the presence of this new type of application impacting material. Unlike open source software, detection mechanisms and evaluation methodologies for open-weight AI models require special treatment, meaning that most present SBOM generation and management tools don't account for them at all.

Regulators are closing that gap: organizations cannot meet compliance requirements unless they track AI models with the same rigor applied to open source software components which minimally means accurate inventory, exposure of metadata, and inclusion in your SBOM output.

In this webinar, we cover what it takes to bring open source AI models under the same governance framework you apply to the rest of your software supply chain:

• What open-weight AI models are, exactly, and how do they relate to the software that ‘runs’ them
• How open source AI models introduce risks (some in common with software, some distinct) that standard SCA tooling wasn't built to catch
• What AI model identifiers and metadata are available during detection, and what they mean in practice
• How to detect AI models early, during build, and surface their presence and characteristics for evaluation before they reach production
• How to extend your existing SBOM and policy-as-code workflows to cover your AI model inventory
• Tools and techniques to evaluate your detected open-weight models against AI model behavior specific anti-patterns (jailbreaks, prompt injection, data exfiltration, toxic payload generation, import hallucinations, and more)