Last updated: March 19, 2026
What is a Transparency Note?
An AI system includes not only the technology, but also the people who will use it, the people who will be affected by it, and the environment in which it is deployed. Microsoft's Transparency Notes are intended to help you understand how the AI technology behind Copilot works, the choices we made that influence the system's performance and behavior, and the importance of thinking about the whole system, so that users of Copilot can take control of their own experiences and understand the steps we are taking to provide a safe and secure product.
Microsoft's Transparency Notes are part of a broader effort at Microsoft to put our AI Principles into practice. To find out more, see Microsoft AI Principles.Â
Overview of AI Restyle in OneDrive
AI Restyle is an artificial intelligence-powered feature designed to creatively transform digital images by applying different artistic or stylistic effects. It leverages advanced machine learning models to analyze input data and generate new versions that reflect selected styles or themes. Key terms associated with AI Restyle include "input content" (the original material provided by the user), "style model" (the algorithm or template applied), and "output" (the transformed result).
How does AI Restyle process my photos? (Cloud vs. device processing)
When you use AI Restyle in OneDrive Photos, the selected photo remains stored in your OneDrive account. To generate the requested edit, a temporary processing copy of the image is securely transmitted to a Microsoft‑managed cloud‑based AI service.
Processing occurs within an isolated execution environment that is scoped only to your specific editing request. The image is not publicly accessible and is not shared with other users or services.
Once processing is complete, the edited result is returned to your OneDrive account for preview and optional saving. The original photo is not modified.
Temporary processing artifacts created during image generation are automatically deleted after the request is completed.
​​​​​​​Use of personal content for model training
Photos processed using AI Restyle in OneDrive are not used to train, improve, fine‑tune, or evaluate Microsoft generative AI models.
Images submitted for editing are processed only to fulfill your specific request and are not added to datasets used to improve or retrain AI systems.
Microsoft does not use personal OneDrive content processed through AI Restyle to develop or improve foundation AI models.
Capabilities
AI Restyle enables users to reimagine content through various creative filters, artistic effects, and style adaptations. It is designed to be intuitive, offering a range of options for users to preview and select preferred outcomes. The system includes safety features such as content moderation, style safeguards, and output review mechanisms to help prevent harmful or inappropriate results. AI Restyle’s core focus is on creative enhancement and expression, rather than realistic edits, factual analysis or critical decision-making.Â
Intended uses and use cases
AI Restyle is intended for scenarios where users wish to enhance, personalize, or share creative content. Example use cases include generating stylized images for social media, customizing personal photos or artwork, and exploring creative inspiration for projects. The system is suitable for personal expression, entertainment, education, and non-commercial applications. Users are encouraged to explore its features while remaining mindful of context and audience.
Considerations and unsupported uses
AI Restyle is not designed for applications requiring factual accuracy, legal or medical advice, identity verification, or critical decision-making. It should not be used to create misleading, harmful, or offensive content. Users must comply with applicable laws and regulations, including copyright and intellectual property rights. The system is not intended for use in high-risk domains, such as law enforcement, healthcare, or financial services, where errors or misrepresentations could have serious consequences.
Limitations
While AI Restyle aims to deliver high-quality creative outputs, it has technical and operational constraints. Generated results may vary in quality, and outputs may not always align perfectly with user expectations or chosen styles. The system may not recognize all cultural or contextual nuances and can occasionally misinterpret input data. Users should review outputs for suitability before sharing and avoid relying on the system for sensitive or consequential tasks.
Are my photos used to train Microsoft AI models?
Photos processed using AI Restyle in OneDrive are not used to train, improve, fine‑tune, or evaluate Microsoft generative AI models.
Images submitted for editing are processed only to fulfill your specific request and are not added to datasets used to improve or retrain AI systems.
Microsoft does not use personal OneDrive content processed through AI Restyle to develop or improve foundation AI models.
System performance
Performance is assessed through a combination of automated testing, user feedback, and quality review. Common error types include style mismatches, incomplete transformations, or unintended artifacts in output. User controls such as style selection, preview, and undo functions are provided to help manage and refine results. Continuous monitoring and updates are in place to address emerging issues and maintain system reliability.
Does Microsoft retain or review my photos after AI editing?
AI Restyle services are designed to minimize retention of personal content.
Full‑resolution image content is not retained by the AI editing service after the processing request is completed. Temporary processing artifacts created during image generation are automatically deleted once the edited result is returned.
To maintain service reliability and detect misuse, Microsoft collects limited service telemetry during processing. This includes usage events, performance diagnostics such as latency and error rates, and safety classification signals. This telemetry does not include full‑resolution image content.
Automated systems may monitor generated outputs to detect harmful content, policy violations, or abusive use. Microsoft does not provide human reviewers access to user photos for routine AI editing operations.
Best practices
-
Clearly specify desired styles or effects for improved results.
-
Review generated outputs carefully before sharing or publishing.
-
Avoid using AI Restyle for sensitive, confidential, or regulated content.
-
Respect copyright and intellectual property when uploading or transforming content.
-
Use human judgment and oversight in all creative workflows involving AI Restyle.
Evaluation methods and resultsÂ
AI Restyle is evaluated using a mix of technical testing, user studies, and expert review. Evaluation focuses on creative quality, user satisfaction, safety, and fairness. Testing includes diverse input scenarios to ensure robustness and minimize unintended bias. Results inform ongoing improvements, and known design constraints are documented and addressed as part of Microsoft’s responsible AI development process.Â
Fairness considerationsÂ
Microsoft is committed to reducing representational harms and promoting fairness in AI Restyle. The system incorporates safety filters to identify and block inappropriate or harmful content. Human review processes are in place to address edge cases and escalate issues as needed. Continuous monitoring helps to identify and mitigate potential biases or unfair outcomes.Â
Integration guidanceÂ
When integrating AI Restyle into applications or services, ensure that users are informed about the system’s AI-driven nature and its limitations. Provide clear user controls, transparency about how outputs are generated, and opportunities for human oversight. Encourage responsible use by outlining appropriate scenarios and discouraging misuse or overreliance on automated outputs.Â
How is my data protected during AI Restyle processing?
AI editing requests are transmitted using encrypted network connections and processed within Microsoft‑managed secure cloud environments that implement industry‑standard security and privacy controls.
Each editing request is handled within an isolated processing context to prevent unauthorized access or data exposure. Temporary processing data exists only for the duration required to complete the requested edit and is automatically deleted afterward.
Generated edits are evaluated using automated safety systems to detect harmful or policy‑violating content and help prevent unsafe outputs. These safeguards operate during the editing process and support fairness and responsible use of generative AI technologies.
Further resources
Feedback and contact
Microsoft welcomes feedback to improve AI Restyle and its associated guidance. To report issues, share suggestions, or request support, please use the in-product feedback tools or visit the Microsoft Support website. Your input helps us advance responsible AI development and user experience.Â
About this document
Copyright © 2026 Microsoft Corporation. All rights reserved. This Transparency Note is provided for informational purposes only and is subject to change. For the most recent updates, visit the official Microsoft documentation.
Published: 01/19/2026
Last updated: 03/19/2026
