Our comprehensive vulnerability detection begins with the industry's most extensive dataset of model security issues. This intelligence is driven by our Huntr community—a network of over 16,000 security researchers who continuously discover novel attack vectors through our bug bounty program and strategic partnership with Hugging Face.

You can explore our most recent discoveries in the Insights DB, where we publish vulnerability assessments from scanning every public model in the Hugging Face repository. The platform includes mechanisms for security researchers and ML practitioners to challenge or validate our findings.

InsightsDB classifies and lists the models as Safe, Unsafe, and Suspicious. You can dispute a finding in Insights DB by selecting the specific model from the list and Report an issue:

• Report your finding if you've found a new threat.
• Report an incorrect threat if you disagree with our findings

This continuous intelligence stream enables us to identify a broad spectrum of vulnerabilities in your models, including many AI-specific security issues that fall outside the scope of traditional software vulnerabilities—threats that typically won't appear in the National Vulnerability Database (NVD) or conventional security feeds.

Model Security operates as a centralized repository of models from third-party sources like Hugging Face.

Additionally, we provide an on-premises scanning solution that deploys directly within your infrastructure, enabling you to assess your proprietary models locally without data transmission to our systems—ensuring complete privacy and security of your assets.