JFrog unveils AI Catalogue to enhance secure model governance

JFrog has announced the launch of its AI Catalogue, a new extension of the JFrog Platform designed to help organisations securely discover, govern, and deploy AI and machine learning (ML) models.

The AI Catalogue is intended to support enterprises in integrating AI services across their software supply chain, providing access to open-source AI models including the NVIDIA Nemotron family. These models are made available with open weights, datasets, and usage guides to help address industry requirements for transparency and control in AI model deployment.

Governance and security

Yuval Fernbach, Vice President and CTO for JFrog ML, emphasised the growing challenges around governance and security as AI adoption accelerates across various sectors. In his statement, Fernbach said:

"One of the biggest challenges for organizations adopting AI is ensuring governance and security to deliver Trusted AI. Building on our Secure Model Registry, the new AI Catalogue provides a centralized hub to access and govern AI/ML models – whether internal, from open-source repositories like Hugging Face, or from external API providers like NVIDIA NIM and Anthropic. By integrating seamlessly with the ecosystem, the JFrog AI Catalogue gives organizations complete visibility, compliance, and control over model usage, helping them innovate faster while delivering Trusted AI in today's complex regulatory environment."

This focus on security and compliance is intended to ease operational complexity and help teams maintain a compliant workflow when introducing third-party or internally developed AI models. The Catalogue is also designed to be scalable, offering continual security scanning and evidence tracking of AI models via JFrog Xray, including model lineage visibility to support documentation efforts and audit trails.

Industry context

Recent research from Gartner has highlighted an increasing need for effective oversight and governance as data science and machine learning initiatives become more distributed within organisations. According to Gartner, "a significant challenge for data science and AI leaders is overseeing and governing the activities of dispersed DSML teams while optimizing collaboration with centralized resources. Enhanced AI governance and management capabilities, linked across data sources and other assets, are now must-have capabilities."

In response, JFrog's AI Catalogue positions itself as a central repository where organisations can locate and manage a variety of AI assets, including models and data sets, with policies and permission controls enforceable on a project-by-project basis.

Capabilities of the AI Catalogue

Key features of the JFrog AI Catalogue include end-to-end model governance, where organisations can track usage and manage access with detailed controls; continuous security integration for ongoing compliance; searchable discoverability via tags and metadata; and the ability to build specialised AI agents. The platform also offers one-click deployment capabilities, whether to the user's own infrastructure or via external AI model providers such as OpenAI and Anthropic.

The Catalogue's integration with open ecosystems covers models from both public repositories and commercial providers, aiming to make it easier for teams to discover secure, production-ready models and deploy them with full visibility.

On the enterprise adoption front, Adel El Hallak, Senior Director of Product at NVIDIA, commented:

"Enterprises face increasing demands for secure, transparent AI model management to maintain compliance and accelerate innovation. By providing direct access to NVIDIA Nemotron models and NIM microservices, within the JFrog AI Catalogue, organizations can deploy and govern open-source AI solutions with greater visibility and control, supporting secure, sovereign AI initiatives across their workflows."

Integration and deployment

The AI Catalogue supports direct integration with external APIs from providers such as AWS, Google, OpenAI, and Anthropic, as well as deployment of containerised models internally. This is designed to simplify the path from AI model discovery to operational deployment while keeping track of usage patterns and compliance requirements.

Teams can now access and manage a wide selection of curated AI models and datasets, improve collaboration between developers and data scientists, and enforce security standards through integrated scanning and tracking features. The ability to centrally govern model access also addresses the need for strong policy enforcement across diverse projects and teams.

The JFrog AI Catalogue is available immediately for users of JFrog Curation, supporting the management of both traditional and AI artefacts.