The document states that fine-tuned models are exclusive to the customer and are not shared with other customers or used to train other models, and that fine-tuning data is retained indefinitely until the customer actively deletes the associated files.
This analysis describes what OpenAI'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 establishes that fine-tuning data does not have an automatic deletion timeline, unlike the 30-day default retention for API inputs and outputs. Customers must actively delete fine-tuning files to trigger removal, creating an open-ended retention period that may require active management for data minimization compliance.
The updated terms state that workspace admins 'can control' data retention rather than directly controlling it. This conditional phrasing may suggest that retention control is optional or contingent rather than a guaranteed capability. Enterprise customers relying on admin-driven data retention policies should clarify with OpenAI whether this change affects their ability to set specific retention timelines for workspace data.
View change record →The updated terms shift governance of conversation access and retention from end users to workspace administrators. Under the revised policy, workspace admins can now view, access, export, and delete any end user conversations within their workspace and control how long workspace data is retained. Additionally, OpenAI now reserves the right to retain deleted or unsaved conversations beyond the standard 30-day deletion window if retention is reasonably necessary to protect its services or any third party from harm, beyond prior language that limited retention extensions to legal requirements. Within an enterprise account, end users no longer have unilateral control over conversation visibility or deletion of their own conversations.
View change record →This provision assures customers that fine-tuned models remain exclusive to them and their training data is retained at their discretion, protecting proprietary customizations and data.
View full change record →Under this provision, fine-tuning data submitted by customers is retained on OpenAI's systems without a fixed expiration date until the customer deletes the files. The agreement states that fine-tuned models are not shared with other customers or used to train other models.
Cross-platform context
See how other platforms handle Fine-Tuning Data Ownership and Retention and similar clauses.
Compare across platforms →"Yes, you can adapt certain models to specific tasks by fine-tuning them with your own prompt-completion pairs. Your fine-tuned models are for your use alone and never served to or shared with other customers or used to train other models. Data submitted to fine-tune a model is retained until the customer deletes the files.Excerpt from OpenAI's Enterprise Privacy
(1) REGULATORY LANDSCAPE: The indefinite retention of fine-tuning data until customer deletion engages GDPR's storage limitation principle, which requires that personal data not be retained longer than necessary.
Enforcement risk, jurisdiction flags, contract triggers, and due diligence action items.
Get the 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.
This provision establishes that fine-tuning data does not have an automatic deletion timeline, unlike the 30-day default retention for API inputs and outputs. Customers must actively delete fine-tuning files to trigger removal, creating an open-ended retention period that may require active management for data minimization compliance.
Under this provision, fine-tuning data submitted by customers is retained on OpenAI's systems without a fixed expiration date until the customer deletes the files. The agreement states that fine-tuned models are not shared with other customers or used to train other models.
No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by OpenAI.