Google permits removal of visible AI watermarks, excluding EU and South Korea
The move balances user flexibility with regulatory mandates, as visible markers are now optional in most jurisdictions but remain compulsory where deepfake laws apply.
Google has announced a significant shift in its approach to AI content labelling, allowing users to remove visible watermarks from generated images and videos. The update, confirmed by Google Vice President Josh Woodward, introduces a toggle feature designed to give creators greater control over their output while maintaining underlying safety mechanisms.
The new capability will initially be available for the Nano Banana, Omni, and Lyria models, which currently utilise a sparkle icon to denote artificial intelligence generation. Google has also confirmed that this functionality will be integrated into its Gemini platform and the Flow video editor at a later date.
Despite the removal of visible markers, the company emphasised that content identification remains possible through technical means. The invisible SynthID watermark and C2PA metadata will remain embedded in all generated files and cannot be disabled by users. These invisible tags require external verification, such as uploading content to a portal or querying a chatbot, to confirm the origin of the material.
The global rollout of this feature will exclude jurisdictions where legislation mandates visible labelling for AI-generated content. This specifically includes the European Union, which recently introduced transparency rules requiring visible watermarks on authentic-looking deepfakes, and South Korea, which implemented similar regulations earlier this year.
Woodward stated that the change aims to strike a balance between creative control and safety. The decision reflects a broader industry trend towards refining AI labelling, with competitors such as Suno and Anthropic also recently updating their transparency tools to comply with evolving regulatory landscapes.
