Notion AI security & privacy practices changed Notion, Oct 19, 2026

1. +Added: At Notion, we want to be transparent with our customers about our products and how we use AI to enhance our users' experience. Below is an overview of Notion AI functionality and related AI governance, security and privacy practices.
2. +Added: Permissions are respected: Notion AI can only ever use content you give it access to.
3. +Added: By default, your customer data is not used to train models.
4. +Added: The current AI provider list with access to Customer Data is always up to date on our Subprocessor page .
5. +Added: Independently certified: Notion AI is in scope for our SOC 2 Type 2 report and ISO 27001 certification and we require equivalent standards from our third-party providers.
6. +Added: Which AI Providers does Notion use?
7. +Added: Notion AI runs on a mix of large language models (LLMs).
8. +Added: Some models are made available to Notion directly by the company that built them; others are served to us through inference providers (for example, a model may be accessed via a specialized infrastructure provider rather than the model's original author). Whether a provider is engaged directly or as an inference provider, all are bound by the same contractual commitments including our no-training and data-handling requirements.
9. +Added: We review providers and models regularly to give you the best experience, so this list changes over time. For that reason, we keep the authoritative, current list on our Subprocessor page rather than here. For each provider, that page shows what their purpose is and where they process data.
10. +Added: See the Data Processing Addendum for details.
11. +Added: When Notion AI answers a question using your workspace, two things happen:
12. +Added: Notion looks through a searchable representation of your pages called an index so it can locate the pages most relevant to your request.
13. +Added: When the user submits a message, Notion creates an embedding for the message so that it can be used to search against the page index.
14. +Added: Generating a response: Those relevant pages and your request are sent to the LLMs, which generate the answer.
15. +Added: Both steps honor your existing permissions and all of the data protections described within.
16. +Added: An embedding is a numerical representation of text that captures its meaning, so pages on similar topics sit close together and can be matched quickly to a request.
17. +Added: Notion generates an embedding for each page using zero-retention LLMs.
18. +Added: Embeddings for each workspace are logically separated each other.
19. +Added: Embeddings are stored in a vector database (currently Turbopuffer) optimized for fast, relevant lookups. Any vector database we use is vetted by our security team and is independently audited against SOC 2 Type 2.
20. +Added: It's passed to the LLMs. If no workspace search is needed, the response is generated and returned. If a search is needed, the LLMs generate a search query.
21. +Added: The query runs against the vector database to find candidate pages.
22. +Added: Those pages are ranked and refined for relevance.
23. +Added: The LLMs generate a response from your request and the ranked pages.
24. +Added: Notion formats the output and displays it to you.
25. +Added: Notion AI respects your existing permissions: The models used to generate a response for you cannot see or use any content you don't already have access to. Notion AI never widens your access.
26. +Added: Always true, for every model: By default, neither Notion nor our AI providers use your Customer Data to train any models. Our contracts with AI providers prohibit them from using your data to train their models, and your use of Notion AI grants Notion no right or license to use your data to train ours. This prohibition on training applies to every provider in the chain: it flows through to both the provider or inference provider that hosts or serves the model, and is enforced through written data-processing agreements with each.
27. +Added: Enterprise customers - Varies by model: Most models we use offer zero data retention (ZDR), meaning they process your data only to generate a response and keep nothing afterward. A small number of models don't offer zero data retention. When you use one of those models, your data may be retained by the provider for a limited period solely to deliver the response. These models are off by default and the workspace owner must opt in to use these models.
28. +Added: Free, Plus, and Business : AI provider retains your data for 30 days or fewer, then deletes it.
29. +Added: There are certain AI-powered features that may require the use of data-retaining LLMs; in such cases, Notion will make available Workspace settings to enable administrators to turn on data-retaining LLMs.
30. +Added: External Agents: Different controls apply. See External Agents .
31. +Added: Delete a page or workspace and you can restore it within 30 days. After 30 days, deleted content is unrecoverable and moves into permanent deletion in line with the retention policies outlined in our Data Processing Addendum. For more information, see: How to delete or restore .
32. +Added: Embeddings are deleted within 60 days of the page or workspace being deleted.
33. +Added: You own your input and output: Notion doesn't claim ownership of what you put into Notion AI or what it generates. As further described in the Notion AI and Notion Credit Supplementary Terms , your Input and Output are your Customer Data.
34. +Added: Rules for using Notion AI: The Notion AI and Notion Credit Supplementary Terms and Notion's Content & Use Policy apply. Violating them can result in removal of content or suspension of access to your workspace.
35. +Added: Encryption in transit: Whenever Customer Data is sent to a provider, it's encrypted using TLS 1.2 or greater.
36. +Added: Customer segregation: Each customer's workspace data is logically separated in production. We do not mix or process workspace data from different customers together. We do not expose your data to other customers.
37. +Added: Provider vetting: Before engaging any subprocessor, we review their privacy, security, and confidentiality practices and enter into a written agreement that imposes data protection obligations consistent with our commitments to you, as described in our Data Processing Addendum .
38. +Added: Further information can be found on:
39. +Added: Notion AI is included in the scope of Notion's SOC 2 Type 2 report and ISO 27001 certification. For workspaces with zero data retention in effect, our providers' zero-retention APIs support HIPAA-compliant workflows.
40. -Removed: Notion offers a suite of AI-powered capabilities, integrated seamlessly across your workspace. Notion AI has the capabilities of multiple AI tools, all in one place. Use it to:
41. -Removed: Notion AI appears seamlessly in your workspace but leverages technology from several AI Subprocessors to provide you with the service. Check out our Subprocessor Page for a complete list of our current Subprocessors, and learn more about Notion AI here →
42. -Removed: Learn more about security features for Custom Agents here .
43. -Removed: Who are Notion's Large Language Model Providers ?
44. -Removed: Notion currently utilizes various large language models (LLMs) hosted by Notion as well as by organizations such as Anthropic and OpenAI. We continuously evaluate LLM providers and their models to provide the highest quality experience to our Notion AI users. Any third parties that process Customer Data will be published in our Subprocessor Page .
45. -Removed: How do I subscribe to new Subprocessor notifications?
46. -Removed: Customers may sign up to receive notification of new Subprocessors by emailing [email protected] with the subject "Subscribe to New Subprocessors." Once a customer has signed up to receive new Subprocessor notifications, Notion will provide that customer with notice of any new Subprocessors before authorizing the new Subprocessor to process Customer Data. For additional information, please see our Data Processing Addendum .
47. -Removed: How does Notion AI use content from a workspace?
48. -Removed: Notion AI is able to reference content from your workspace through two key phases:
49. -Removed: Generating an embedding (i.e., a semantic representation of the information stored in your workspace).
50. -Removed: Generating responses informed by the content in your workspace.
51. -Removed: Embeddings are numerical representations of text or documents. These representations capture the meaning and context of the text in a multidimensional space, where similar topics have similar numerical representations. By using embeddings, vector search algorithms can efficiently compare and find similarities between different pieces of text or documents. For Notion AI, embeddings are created from workspace content to enable the system to provide accurate and relevant responses to user questions.
52. -Removed: Here is an example of an embedding from OpenAI :
53. -Removed: For each page in your workspace, we generate an embedding by using an OpenAI zero-retention embeddings API.
54. -Removed: Notion receives an embedding for each Notion page and stores it in a vector database (e.g., Turbopuffer). The vector database is a data store optimized for embeddings that enables fast lookup of the most relevant pages given a user request.
55. -Removed: Notion treats embeddings with the same level of security and privacy considerations as Customer Data. All our Customer Data commitments outlined in our Master Service Agreement (MSA) and Data Processing Agreement (DPA) apply to embeddings. View our Terms and Privacy Page for more information.
56. -Removed: We store embeddings with vector databases like Turbopuffer. These vector databases have been vetted by our security team as well as an external auditor to obtain their SOC 2 Type 2 certification.
57. -Removed: Notion receives a request from a user.
58. -Removed: The request is passed to LLMs and AI Models. If the request does not require searching the Notion workspace, a response is generated at this point and returned. If the user's request requires searching their workspace, LLMs and AI Models generate a search query most relevant for the user request.
59. -Removed: The query is passed to a vector database, where a list of pages is found based on relevance to that query.
60. -Removed: Notion sends the query - and the pages identified by the vector database - to LLMs and AI Models where the pages are refined and ranked by relevance to the query.

33 more changed paragraphs are on the page itself.