Legal 365 Snapshot: AI Watermarks and Metadata Coming to Microsoft 365

Starting in late February 2026, Microsoft is introducing a new Cloud Policy setting called “Include a watermark when content from Microsoft 365 is generated or altered by AI.”  

When enabled, this policy adds visible watermarks to AI-generated or AI-altered video content and audible watermarks to AI-generated or AI-altered audio content across Microsoft 365.  Think Clipchamp-generated videos or Copilot audio overviews created from Word documents. 

🔗Microsoft 365 Roadmap – Feature #554934

This is an admin-controlled policy managed through Cloud Policy service for Microsoft 365.  It defaults to Disabled/Not Configured, meaning no watermarks are applied unless the tenant admin affirmatively turns it on.

What You Need to Know

• Opt-in, not default.  Organizations must deliberately enable this policy.  If you do nothing, no watermarks are added to video or audio content. 
• No customization.  You cannot modify the placement or wording of watermarks.
• Not available for government tenants.  This policy is not available to U.S. government customers using GCC, GCC High, or DoD offerings, which is a gap for public sector organizations. 
• Metadata for video and audio is not yet available.  Microsoft states that regardless of whether you enable the watermark policy, provenance metadata aligned with C2PA (Coalition for Content Provenance and Authenticity) standards will be added to AI-generated or AI-altered video and audio content in the future.  Currently, however, this metadata capability is only available for AI-generated images and is not yet functional for video and audio.  Microsoft says this is in development but with no specific timeline.  (More on this below.)
• Defensibility matters.  Enabling watermarks creates a visible, auditable signal on AI-generated video and audio content.  Currently, only certain jurisdictions require AI-generated content to be identifiable, and the requirements vary between mandatory watermarking/metadata, visible labeling, and context specific disclosure.  (See, e.g., Article 50 of the EU Artificial Intelligence Act.)  Elsewhere, failure to visibly disclose content as AI-generated could be viewed as misleading and run afoul of consumer protection, election, or deepfake laws and regulations.  As regulators worldwide increase scrutiny of synthetic media and AI-generated deepfakes, ensuring that the outputs of AI systems are marked in a machine-readable format and detectable as artificially generated or manipulated may carry weight in litigation or regulatory inquiries. 
• Metadata ≠ watermark.  Don’t conflate the two.  Metadata is invisible to the average viewer and can be stripped from files.  A visible or audible watermark is a far more durable signal of AI origin. 

The Image Exception

The Cloud Policy described above does not apply to images.  Instead, Microsoft is putting image watermarking in the hands of individual users, who can toggle it on or off through their own privacy settings at myaccount.microsoft.com (anticipated second half of February 2026).

This means there is no admin-level policy to enforce watermarking on AI-generated or AI-altered images across the organization.  If a user doesn’t opt-in, images created through Microsoft Designer or the AI image generation features in Word and PowerPoint will not carry a visible watermark, though provenance metadata will still be added to these images.    

For organizations that need consistent, enforceable labeling of AI-generated images, this is a governance gap.  Your practical options right now are:

• Disable AI image generation entirely using the “Control access to Designer Image Generation” policy in Cloud Policy, which affects Microsoft Designer and image generation in apps like Word and PowerPoint. 
• Issue clear internal policies and guidance requiring users to enable image watermarks, while acknowledging you cannot technically enforce it at the tenant level. 

It is important to note that some of this functionality has not yet rolled out and could be implemented differently once released fully to M365 tenants.  It is very important to test these features prior to releasing them to Enterprise M365 tenants. 

🔗Microsoft 365 Roadmap – Feature #554934

Legal 365 for Microsoft 365

Redgrave’s Legal 365 Snapshots provide timely thought leadership and practical insights, which highlight the importance of a combination of deep technical and legal insight when managing Microsoft 365 environments.  Learn more about our Legal 365 for Microsoft 365 services, which help organizations confidently navigate complex M365 governance challenges with defensible, efficient, and legally sound strategies, here.

The views expressed in this article are those of the authors and do not necessarily represent the views of their law firm or any of its clients.