How can I request data deletion from OpenAI? What does the process entail?
This fact-check may be outdated. Consider refreshing it to get the most current information.
Executive summary
OpenAI provides multiple paths for individuals to request deletion of personal data: account-level controls in ChatGPT, a Privacy Portal where users can “Make a Privacy Request,” and region‑specific Personal Data Removal Request forms used especially in Europe and Japan (OpenAI Help Center; TechCrunch) [1] [2]. The process can remove chat history and stop future use of a person’s messages for training, but it may take time, require identity verification, and does not clearly guarantee removal from the models’ underlying training datasets [3] [2] [4].
1. What tools OpenAI publishes for deletion and opt‑out
OpenAI’s published routes begin with in‑app Data Controls — users can disable chat history and toggle “Do not train on my content” in ChatGPT — and extend to the Privacy Portal where a “Make a Privacy Request” button feeds formal deletion or account‑deletion workflows (OpenAI Help Center; CNET) [1] [3]. For Europeans and Japanese users there is an additional “Personal Data Removal Request” web form designed to exercise regionally protected rights; TechCrunch and other reporting identify that form as the primary mechanism to request that OpenAI remove personal data or stop processing it for model training [2] [5].
2. What users will be asked to provide
The Privacy Portal and regional removal forms expect identifying information and sometimes corroboration that the individual is mentioned in model outputs; reporting notes that OpenAI asks for account details when available and in some cases requests sworn accuracy statements and supporting evidence for claims that the model has output personal information (TechCrunch; Analytics Vidhya) [2] [5]. Community reports and help docs indicate OpenAI may request additional verification to ensure the request is legitimate before acting [6] [7].
3. What OpenAI says it will do and expected timelines
OpenAI’s Help Center says account deletion initiated through the Privacy Portal or ChatGPT will lead to deletion of data within about 30 days, though the company may retain a limited set of records where required by law (OpenAI Help Center) [1]. CNET and other guides explain that disabling training stops future use of new conversations for model training but that OpenAI retains disabled chats for a 30‑day abuse‑monitoring window before permanent deletion [3].
4. Limits, caveats, and the contested question of training data
Independent reporting stresses important limits: many deletion requests target content that appears in ChatGPT responses or user‑account data, but whether and how an individual’s information can be removed from the underlying model weights or third‑party backups is unclear and contested (Wired; TechCrunch) [4] [2]. Journalistic and community sources warn OpenAI balances “privacy and free expression” when evaluating removal requests, and that systemic removal from models trained on large-scale scraped data is a harder technical and policy problem than erasing account history [5] [4].
5. Friction, dark patterns, and enforcement dynamics
TechCrunch reports the in‑product route to opt out contains user‑interface frictions and “dark patterns” that may discourage users from fully disabling training, and regulators in Europe have leveraged data requests in enforcement actions—so how OpenAI responds can trigger further oversight [2] [4]. Advocacy groups and journalists framing this debate highlight the company’s dual incentives: reducing regulatory risk by offering deletion paths while retaining model training feedstock and minimizing operational burdens [2] [4].
6. Practical step‑by‑step summary
The clearest, documented path is: use ChatGPT’s Data Controls to disable chat history and training for new content; if account removal is desired, submit an account deletion request via OpenAI’s Privacy Portal (“Make a Privacy Request”) or via the in‑app deletion workflow; if seeking removal of specific mentions in outputs or to exercise regional rights, use the Personal Data Removal Request form for Europe/Japan and be prepared to supply identity verification and evidence for the claim (OpenAI Help Center; CNET; TechCrunch) [1] [3] [2]. Reported timelines are on the order of weeks for account deletion and 30 days retention for monitoring prior to permanent deletion, with possible longer retention where legally required [1] [3].
7. What reporting doesn’t fully answer
Available sources document the interfaces and promise of deletion but do not provide independent, verifiable proof that data is expunged from all model artifacts or third‑party backups; OpenAI’s public guidance and journalistic coverage acknowledge that complete removal from pretrained model parameters remains a complex, under‑resolved issue [4] [2]. Therefore, expectations should be calibrated: account data and chat histories can be deleted and future training opt‑outs respected, but erasing influence from historical model training is not definitively demonstrated in the public record [2] [4].