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 prohibits using Core Models, Derivative Works, or outputs generated from them to create new model architectures that compete with Stability AI's Core Models, while permitting the creation of fine-tunes, LoRAs, hypernetworks, and retrains.
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 restricts a specific category of AI development activity (creation of new foundational model architectures) for all Community License users, including researchers and commercial developers, and may affect AI research workflows and downstream product development pipelines.
Interpretive note: The document does not provide a precise technical or legal definition of 'new model architecture,' creating interpretive ambiguity for organizations engaged in large-scale model development or research.
Under the Community License, users may create fine-tunes, LoRAs, hypernetworks, and retrains, but the agreement states they may not use Core Models, their outputs, or Derivative Works to create new foundational model architectures. This restriction applies to both non-commercial and commercial users operating under the Community License.
Cross-platform context
See how other platforms handle Prohibition on Competing Foundational Models 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 20 platforms.
"The Community License places restrictions on creating competing 'foundational models' using the Core Models, Derivative Works and their outputs. 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 intellectual property law, specifically copyright and sui generis database rights to the extent they apply to AI model weights and architectures. The EU AI Act may interact with this provision in contexts where foundational model development is subject to general-purpose AI model obligations. No specific data protection regulations are directly implicated. (2) GOVERNANCE EXPOSURE: Medium. Organizations conducting AI research or model development using Stability AI's Core Models must evaluate whether their activities constitute creation of a new foundational model architecture under the agreement's definitions. The document distinguishes fine-tunes from foundational models but does not provide a precise technical definition of 'new model architecture,' creating interpretive uncertainty. (3) JURISDICTION FLAGS: Enforceability of restrictions on outputs may vary by jurisdiction. In some EU member states and the UK, the extent to which a licensor can contractually restrict use of model outputs (as opposed to the model itself) may engage copyright and database right frameworks. EU AI Act obligations for general-purpose AI models may interact with this restriction for organizations developing large-scale derivative models. (4) CONTRACT AND VENDOR IMPLICATIONS: B2B contracts involving sublicensing or redistribution of models derived from Stability AI Core Models should include representations confirming that downstream development does not constitute foundational model creation. Research agreements with academic or commercial partners using these models should be reviewed against this restriction. (5) COMPLIANCE CONSIDERATIONS: Legal teams should establish clear internal definitions distinguishing permitted fine-tuning activities from prohibited foundational model creation, and document the technical basis for that classification. Organizations engaged in large-scale model research should obtain written clarification from Stability AI if their development activities approach the boundary of this restriction.
This provision restricts a specific category of AI development activity (creation of new foundational model architectures) for all Community License users, including researchers and commercial developers, and may affect AI research workflows and downstream product development pipelines.
Under the Community License, users may create fine-tunes, LoRAs, hypernetworks, and retrains, but the agreement states they may not use Core Models, their outputs, or Derivative Works to create new foundational model architectures. This restriction applies to both non-commercial and commercial users operating under the Community License.
No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by Stability AI.