Google signs EU code on transparency for AI-generated content

Google has signed the European Union’s Code of Practice on Transparency of AI-Generated Content, tying one of the company’s public AI policy commitments to the growing effort to help people understand how online material was made or altered. The move places Google among the companies working with European rules that focus on disclosure, provenance and the practical presentation of information about AI-generated content.

The code sits alongside the wider EU AI Act framework and concentrates on the transparency questions that arise when synthetic text, images, audio or video move through public platforms. The core challenge is easy to describe but difficult to solve at internet scale: people need useful context about content without being confronted with labels so numerous, vague or inconsistent that they stop conveying meaning. A credible system has to work across products and platforms rather than only inside a single company’s tools.

Google said its participation builds on an earlier commitment to the general-purpose AI code of practice and on work connected to the C2PA technical standard. C2PA is designed to carry provenance information about digital content, while Google’s own SynthID technology is used to embed signals that can help identify material created or edited with AI. The company has also pointed to collaboration with other AI and technology groups as part of the push for interoperable approaches to watermarking and disclosure.

For people using online services, the practical value of such measures will depend less on the name of a standard than on whether the information is visible and understandable at the moment it is needed. A viewer deciding whether to trust an image, a publisher checking material before distribution or an advertiser reviewing creative work may all need different kinds of context. The design of labels, the ability of services to preserve provenance data and the treatment of content that has been edited multiple times will therefore matter as much as the underlying technical marker.

Google’s announcement also made clear that it sees a risk in layering too many overlapping legal notices and technical labels onto the same content. That concern is not an argument against transparency; it is a reminder that disclosure systems must be legible if they are to support informed decisions. The company said technical solutions are still evolving, and it urged implementation that gives people clear context rather than a confusing collection of signals.

The policy step arrives while companies, regulators and standards bodies are still working out how disclosure obligations should function across borders and across different types of media. Watermarks and provenance records can help establish a trail, but they do not answer every question about accuracy, intent or the reliability of a claim in the content itself. Users will still need to evaluate the material they see, and platforms will still need rules for harmful or deceptive uses that go beyond a label.

Google did not announce a single new consumer-facing feature or a universal timetable for every product as part of the signing. The immediate significance is instead the company’s agreement to participate in a shared transparency framework while continuing to develop its own tools. The next test will be implementation: whether the standards, product interfaces and regulatory guidance can give people more useful information about AI-generated content without making the online experience harder to understand.