Trust in digital content is becoming a key issue in the age of AI. So-called Content Credentials could help address this challenge in the future. Berlin-based startup Valid Technologies has developed Lumid, a solution centered around these digital markers that enables companies and creators to make the origin, editing history, and AI involvement in media content transparent. The company recently reached an important milestone: Valid is now officially authorized to label content with Content Credentials according to the C2PA standard — the first company in Germany to do so.

We spoke with co-founders Hans and Jens Brorsen and Daniel Larena about what this means for them and why transparent provenance information is becoming increasingly important online.

Hi Daniel, Hans and Jens! First of all, congratulations! Valid has recently received official approval to label media content with Content Credentials. Was it a long road to get there?

“Thank you very much! We’re very proud of this positive outcome and of the trust placed in us by the responsible standards body, C2PA. Even before starting the conformity assessment — as the approval process is referred to in C2PA terminology — we had already gained experience in labeling content, especially images. Still, the road involved a number of organizational and technical hurdles, largely because the review process is taken very seriously. That level of rigor meant a lot of work for us, but it’s also a major quality feature of the standard and part of what makes it trustworthy.”

For readers who may not be familiar with the term: What exactly are Content Credentials, and how do they differ from traditional watermarks?

“You can think of Content Credentials as digital passports embedded directly into media content. They contain machine-readable information about where a piece of content originally came from, how it has been modified, whether AI was used, who owns the rights, and under what conditions the content may be reused or used to train AI systems. One major advantage is that more and more platforms and applications are making this information visible to users. On LinkedIn, for example, images and videos linked to Content Credentials display a CR logo in the upper left corner. Clicking on it reveals additional information about the content’s origin and editing history. Other major tech companies such as OpenAI, Google, Meta, and ByteDance are also working to make Content Credentials visible on their platforms. In services like ChatGPT, Gemini and YouTube, this is already live; integrations have also been announced for Google Search, Chrome, Facebook, Instagram and TikTok.

And how do they differ from conventional watermarks?

The biggest difference is interoperability. Traditional watermarks can certainly be a useful indicator, but they lack a shared language. Content Credentials go several steps further by using a common cryptographic standard adopted by platforms, AI providers, and content creators. This enables secure, cross-platform content verification for users.”

You are the first company in Germany to successfully complete the C2PA process — and you’re a startup on top of that. What does that mean to you: more of a technical milestone, a market signal, or a political statement?

“All three, really. First and foremost, it’s a technical milestone: as a startup, we’ve shown that we can meet the requirements of an international standard like C2PA. Technical excellence has been our ambition from day one, so being able to prove that capability visibly now is very important to us. At the same time, it’s a clear market signal: media consumption and media production have evolved globally, especially with the rise of AI. Content can now be created faster, altered more easily, and classified less reliably. German companies, institutions, organizations, media professionals, and creators need to keep pace. Content Credentials play an important role here, which is why we strongly encourage their adoption. And yes, it’s also a political statement. In times of AI-generated or AI-manipulated media, transparency has become a quality feature. Content Credentials can signal that transparency to both humans and machines.”

Let’s take a step back. Digital disinformation has been around for decades, but AI has significantly intensified the problem. At the same time, trust in reliable content is declining. When did the idea for Lumid emerge?

“The original idea for Lumid emerged before the age of AI. With the rise of platforms and social media, content was suddenly no longer consumed at the source, but somewhere downstream. That opened the door to manipulation and triggered the first major wave of digital disinformation. We wanted to counteract exactly that development from the very beginning.”

Do you feel that awareness of deepfakes and manipulation is growing at the moment? Or are media and society still not taking these issues seriously enough?

“More and more people are concerned about this development, and according to studies, fear of digital disinformation is currently seen as one of the major threats of our time. This is also reflected in regulations such as the Digital Services Act and the EU AI Act. Still, we believe that traditional media, public authorities, and policymakers need to do more. The issue needs to be addressed more concretely, and technical tools to combat it need to be deployed as quickly as possible.”

With Lumid, you’re not focusing on forensic detection of manipulated content after the fact. Instead, you actively mark authentic content. Are both approaches equally useful?

“That’s true, but the important point is this: it’s not just about labeling ‘real’ content. Content Credentials can also be used to transparently label AI-generated or AI-edited content. They can disclose whether content comes from a reliable source and whether AI played a role in its creation. Those two things are not mutually exclusive today. Forensic methods still remain important. They are especially useful where content has not been labeled — for example in fraud cases or manipulated media content. The downside of these so-called detection techniques, however, is that they can be inaccurate and don’t always scale reliably. We approach the issue from another angle: Content Credentials provide provenance and editing information directly with the content itself. This information can be viewed by people and automatically read by machines. That makes the approach particularly scalable — provided that as many platforms, software providers, AI systems and media companies as possible support the same standard. In our view, both are essential: active labeling through standards like C2PA, and complementary forensic methods for dedicated use cases.”

You combine the C2PA standard with proprietary fingerprinting technology. How should one imagine that — and do you really need both?

“Belt and braces! With the C2PA standard, relevant information such as the source, the use of AI, edits, and licensing details is cryptographically signed and written directly into the carrier file itself — for example, a JPEG image file. However, if that file is changed, say by being sent via WhatsApp or turned into a screenshot, that information may be lost or become unreadable. Our fingerprinting mechanism offers a way to recover the original image together with all the relevant information. You can think of it as a kind of safety net.”

When you speak with potential customers — companies, public authorities or platforms — where are you currently seeing the greatest pressure?

“That varies. The C2PA standard is spreading rapidly, and there is strong interest in the creator economy in using Content Credentials to signal trustworthiness across platforms and improve discoverability. The same applies to large companies and public authorities, which are additionally concerned about false information being spread about them — for example through fake profiles or social engineering.

At the same time, we also see considerable interest in using Lumid to comply with the requirements of the AI Act. From August 2, 2026 onward, all AI-generated or AI-manipulated content published in a professional context must be disclosed accordingly. That can be done using Content Credentials. Similarly, AI providers must label their content in a machine-readable way. Lumid can do that as well.”

Valid was founded in 2023, so it’s still a young company operating in a highly competitive technology market. Were there moments when you doubted the decision to start the company?

“Of course there were. Anyone who starts a company — especially in a dynamic and highly competitive tech market — goes through phases where things feel uncertain or don’t move as quickly as one would hope. What matters to us is how we deal with those moments. From the very beginning, we believed that trust, transparency, and provenance for digital content would become massively more important. That conviction carried us through difficult phases, and in hindsight, those very moments were important because they forced us to constantly improve our product, our positioning, and our standards.”

How helpful were the regional ecosystem in Potsdam and the MTH Accelerator in supporting your growth?

“The regional ecosystem, including the MTH Accelerator and also your neighbors at the Media Innovation Center Babelsberg, has been incredibly valuable to us. Especially as a young company, it helps enormously to be in an environment where innovation and media practice come together so directly, and where you can quickly gain access to relevant contacts, expert sparring, and industry knowledge. We’re grateful to be part of this ecosystem and are delighted to continue working with you as alumni.”