Content usage and handling

Last updated on Aug 21, 2025

Get answers to frequently asked questions about Adobe’s content usage and handling practices specific to improving generative AI capabilities in Acrobat apps.

For purposes of the following FAQs, references to “generative AI capabilities” are specific to the AI Assistant and the Generative summary tools only.

Refer to the content analysis FAQ to learn how Adobe may analyze metadata from your prompts (such as the type of question) and documents (like page count, structure, type, and other statistics) for product improvement purposes.

For details on Adobe's security, privacy, and AI ethics practices, visit the Adobe Document Cloud security page, explore resources in our Trust Center, the Adobe Privacy Policy, and the Adobe AI Ethics page, respectively.

Adobe does not look at your documents, prompts, or generated responses except in the following instances:

  • Reported content, bugs, or vulnerabilities: When you report content as harmful, illegal, offensive, or other issues, we investigate by manually reviewing the document, prompts, and generated responses to adjust the service to address the issue.
  • User feedback: For Acrobat and Adobe Scan Individual users* that provide feedback, you have the option to share with us your document, prompts, and generated responses during a document session for product improvement purposes that do not include training a Large Language Model (LLM). Examples of product improvement include improving the operability of generative AI in Acrobat apps, as well as reducing hallucination, bias, and toxicity. If you do not wish to share your content, please uncheck the product improvement checkbox when you first provide feedback on a document.

If manual review of your content takes place, a limited group of trained Adobe personnel will examine it within an encrypted repository with access controls.

When you share your documents, textual prompts, and generated responses with us when providing feedback, we filter out personal information by applying data masking procedures before using the content for product improvement. We replace personal information with predefined categories using Named Entity Recognition, for example, replacing “John Smith” with “PERSON.”

Content is stored for 30 days in an encrypted, siloed, Adobe-controlled environment within the US with strict access controls. You can always contact us via the form in our Privacy Policy if you want us to delete this data sooner, or if you want to opt out of the use of your content to improve generative AI capabilities in Acrobat apps.

The flagged content is only retained for 30 days unless we have other legal reasons to retain its access as described in our Privacy Policy. You can always contact us via the form in our Privacy Policy if you want us to delete this data sooner.

When you open a PDF and select AI Assistant in Acrobat or Acrobat Reader, Generative AI analyzes your document using Adobe servers. The processed data is temporarily cached for up to 12 hours. However, in the desktop version of Acrobat and Acrobat Reader, the AI-generated overview that appears when you open a PDF is processed locally on your device, not on Adobe servers.

You can use AI Assistant for your first five questions, with or without signing in. To continue beyond the initial limit, you must be signed in and subscribed to a paid plan. On Adobe Scan, however, sign-in is required from the start to use the generative AI features.

Note

The generative AI features are only available in English for Individual or Teams users. Enterprise users may only receive generative AI features when their admin assigns them a paid or trial license, or if they are accepted into one of the programs in our private beta.  

You can turn off generative AI to prevent documents from being processed by the feature. 

With AI Assistant in Acrobat, Acrobat Reader, and Adobe Scan, Adobe takes an LLM-agnostic approach, selecting best-in-class technologies that address a range of customer use cases. We have currently integrated the Microsoft Azure OpenAI Service with our proprietary technologies to provide generative AI capability in Acrobat apps.

No. Adobe does not train any LLMs on your content concurrent to your interactions with generative AI in Acrobat. Prompts provided to AI Assistant do not modify the underlying model. Your content will not be used to train any LLMs that deliver Acrobat’s generative AI capability. 

Acrobat AI Assistant supports Regional Data Center Pinning for enterprise customers in the European Union (EU) and United States (US), ensuring that user data is processed and temporarily cached (for up to 12 hours) within their designated region. For EU Enterprise users, data is cached using Adobe cloud storage in EU (Ireland), and LLM processing takes place in the EU (currently in Sweden). Learn more about Document Cloud Data Centers

Note

Adobe does not use content from an organization or school account for product improvement for generative AI unless otherwise agreed to by the organization or school.

Chat history data is stored on your local device for desktop and mobile apps and in Adobe cloud storage for Acrobat on the web. It includes your textual prompts and generated responses for a specific document or a set of documents you selected, also termed as a Collection. To access it, open the relevant file or Collection from your recent files list.

The chat history might still contain content from documents you've removed from a Collection. The AI Assistant may also use this chat history as context when responding to your prompts.

No, chat history can't be exported or shared. You can manually copy responses if needed.

Select More > Clear chat history from the AI Assistant panel of a document or Collection to delete all prompts, responses, and overviews.

End users can't disable chat history, but enterprise admins can do so via the Adobe Admin Console.

You’ll lose access to related files and their associated chat history in Acrobat.