TL;DR: Microsoft is introducing a new administrative policy to add visual and audio watermarks to AI-generated content. Available via Cloud Policy, this feature targets video and audio, while images remain a user-level setting. This should be rolling out in February according to Microsoft.
The Need for AI Transparency
As generative AI becomes a staple in the modern workplace, the line between human-made and synthetic content is blurring. For IT Managers and Compliance Officers, the risk of ‘ deceptive content ‘ is a growing concern. Microsoft’s new watermarking policy is designed to provide clear, undeniable signposts for AI-altered media, aligning with the Microsoft Enterprise AI Services Code of Conduct.
How the New Policy Works
The new setting, titled ‘ Include a watermark when content from Microsoft 365 is generated or altered by AI ‘, is managed exclusively through the Cloud Policy service for Microsoft 365.
When Enabled, the system applies specific markings:
- Visual Watermarks: These appear on videos generated or edited using tools like Clipchamp.
- Audio Watermarks: These are embedded into audio files, such as those generated by Copilot when creating summaries or overviews from Word documents.
Administrative Setup
Because this feature is disabled by default, organizations must take proactive steps to implement it. It is important to note that you cannot currently customize the placement or wording of these watermarks, they are standardized by Microsoft to ensure universal recognition.
Metadata vs. Visual Markings
A common question for IT teams is what happens if the policy is left ‘ Not Configured ‘. Microsoft has confirmed that even without a visible or audible watermark, metadata regarding the AI’s role in content creation is still added to the file. This ensures that while a casual viewer might not see the mark, forensic or audit tools can still identify the content’s origin.
The Image Watermark Split
One nuance of this rollout is the handling of images. Unlike video and audio, which are controlled via central admin policy, image watermarking is currently a user-level privacy setting.
- Where to find it: Users must navigate to
Settings & Privacy > Privacyatmyaccount.microsoft.com. - The Impact: Once enabled by the user, images generated in Word, PowerPoint, or Microsoft Designer will carry a visual watermark.
Strategic Impact for Organizations
For companies operating under the EU AI Act, these features are vital. The act requires that synthetic content be clearly labeled to prevent the deception of users. By enabling this policy, IT departments provide a safety net for their users, ensuring that the organization remains compliant with emerging international AI regulations.
FAQ
When will this policy be available for my tenant? This should be rolling out in late February 2026 according to Microsoft.
Does this watermark appear on text generated by Copilot? No. This specific policy is focused on rich media. Text-based watermarking or ‘ digital fingerprinting ‘ for Word or Outlook text is not part of this roadmap ID.
Can a user remove the watermark? The watermark is baked into the exported file. While sophisticated users might attempt to edit it out, the underlying metadata remains as a secondary layer of proof regarding the content’s AI origin.
Is there an extra license required for watermarking? No, this is a policy setting available within the Cloud Policy service for standard Microsoft 365 commercial tenants.
Conclusion
The introduction of automated watermarking is a significant step toward a transparent AI workspace. By centralizing these controls, Microsoft helps organizations meet legal requirements while maintaining trust in digital assets.