Get the weekly research letter
Companies change their terms quietly. We read every version and catch what actually changed. One email a week on the changes that matter and what they mean. No account.
The agreement permits the creation of fine-tunes, LoRAs, hypernetworks, and retrains from Core Models, and permits using outputs from these Derivative Works to train further derivatives, but prohibits using Core Models or their outputs to create new model architectures.
This analysis describes what Stability AI's agreement states, permits, or reserves. It does not constitute a legal determination about enforceability. Regulatory applicability and practical outcomes may vary by jurisdiction, enforcement context, and individual circumstances. Read our methodology
This provision defines the boundary between permitted derivative model development (fine-tuning) and prohibited foundational model creation, which has direct implications for AI development workflows and downstream IP arrangements.
Interpretive note: The document does not provide a precise technical definition of 'new model architecture,' which may require case-by-case assessment for large-scale or structurally modified derivative models.
Under these terms, users may freely create and commercialize fine-tunes, LoRAs, and similar Derivative Works from the Core Models, and may use the outputs of those derivatives to train further models. The agreement prohibits using Core Models or their outputs to create new foundational model architectures.
Cross-platform context
See how other platforms handle Fine-Tune and Derivative Work Permissions and similar clauses.
Compare across platforms →Monitoring
Stability AI has changed this document before.
Receive same-day alerts, structured change summaries, and monitoring for up to 25 platforms.
"LoRAs, hypernetworks, fine-tunes, retrains, etc, are not 'foundational models'. If you have created a LoRA, you can use its outputs to train other LoRA's. You can't use a Core Model, or its outputs, to create a 'new' model architecture.Excerpt from Stability AI's Model License
(1) REGULATORY LANDSCAPE: This provision engages IP law applicable to software and model derivatives. In the EU, the treatment of AI model fine-tunes and their outputs under copyright and database right frameworks is an evolving area. The EU AI Act may impose additional transparency and documentation requirements on organizations deploying fine-tuned models in high-risk applications. (2) GOVERNANCE EXPOSURE: Low to Medium. The permission to create and commercialize fine-tunes under the Community License is relatively clear, but the prohibition on foundational model creation requires technical and legal assessment of where fine-tuning ends and new architecture development begins. (3) JURISDICTION FLAGS: Organizations in the EU and UK should assess whether fine-tuned models derived from Stability AI Core Models are subject to general-purpose AI model obligations under the EU AI Act, depending on the scale of deployment and capabilities of the resulting model. (4) CONTRACT AND VENDOR IMPLICATIONS: Developers selling or licensing fine-tuned models derived from Core Models should include appropriate representations in their downstream agreements regarding the license basis for the derivative model and any restrictions that flow through from the Community License terms. (5) COMPLIANCE CONSIDERATIONS: Development teams should document their technical classification of derivative model development activities to establish that fine-tuning activities fall within the permitted category and do not constitute foundational model creation under the agreement's terms.
Full institutional analysis
Regulatory citations, enforcement risk, and due diligence action items.
Monitor: same-day alerts on the platforms you choose. Analyst: full institutional analysis.
Compliance Governance Intelligence
Need to monitor specific governance provisions?
Compliance includes provision-level monitoring, governance timelines, regulatory mapping, and audit-ready analysis.
Built from archived source documents, structured governance mappings, and historical version tracking.
This provision defines the boundary between permitted derivative model development (fine-tuning) and prohibited foundational model creation, which has direct implications for AI development workflows and downstream IP arrangements.
Under these terms, users may freely create and commercialize fine-tunes, LoRAs, and similar Derivative Works from the Core Models, and may use the outputs of those derivatives to train further models. The agreement prohibits using Core Models or their outputs to create new foundational model architectures.
No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by Stability AI.