In the face of the flood of AV content created with AI — and not always with good intentions — is there any formula to distinguish what is authentic from what is not?

*This content is included in the February issue of TM BROADCAST. You can access the full magazine here.

To the same extent that artificial intelligence tools have been massively deployed, and their results in the audiovisual field have improved at a dizzying pace, the creation of false news or “fake news” for all kinds of purposes and intentions has also spiraled out of control.

This very concept, which was never a problem in cinema because we always knew that our blue-caped superhero did not really fly (sorry to reveal the secret if anyone did not know…), now represents an extremely complex challenge in the world of journalism and news.

Although the trained eye is still capable of distinguishing at first glance between real content and “artificially created” content, a significant portion of the audience can no longer tell the difference. And at the current pace of development, we predict that in a very short time — a few years at best — it will be practically impossible to distinguish, even for the most experienced eyes.

The first issue we face, being rigorous and as mere spectators, is the paradigm shift this change of scenario represents. We come from a world in which an image was a document in itself and was worth more than a thousand words, only to plunge abruptly into the confusion of not being able to trust, a priori, any of the content before our eyes.

And this brings to the surface what we believe to be the next major challenge for all broadcasters: how to maintain editorial prestige and preserve audience trust. Let us not lose sight of the fact that a single lie can call into question the credibility earned over many years of serious and committed work.

For this reason, we believe that the next major challenge no longer lies in mere technological evolution. It is no longer about HD, UHD, HFR, BT2020, IP… but about ensuring that the information we deliver to our audience is sufficiently trustworthy to maintain our position as credible informers. This challenge is far from easy, nor is it something that can be developed as an individual initiative. Fortunately, and with this objective in mind, the C2PA consortium was created.

The acronym stands for Coalition for Content Provenance and Authenticity. Its purpose is to develop an open technical standard that makes it possible to create and manage content credentials that allow the origin to be identified and all modifications undergone by digital audiovisual content to be tracked.

The overlay that we will likely begin to see soon — and increasingly frequently — in validated content will be “content credentials” or “cr” in its abbreviated version.

It is a technology very similar to what has been available in still photography for some time, but in this case applied to content of much greater length and complexity, such as video and audio. And making it open facilitates the participation of the greatest possible number of stakeholders across all communication and distribution sectors, although it also means that certain developments and implementations require more time for agreements and consensus.

The next major challenge no longer lies in mere technological evolution, but in ensuring that the information we deliver to our audience is sufficiently trustworthy

Although it is not yet fully developed and finalized, we are already in a sufficiently advanced phase of development and demonstration to have it accessible and to assess the change it may represent, both in creative processes and especially in those related to news production. The path has been laid out.

Conceptually it seems simple: it consists of accrediting the origin by stamping a digital signature or certificate in the non-editable metadata of the content during its creation, and progressively adding signatures or certificates that identify what changes have been made to that content until it reaches us.

And this, which stated this way sounds very simple, is extremely complex to put into practice. It must cover the entire creative process: from content capture to viewing by the final audience. Passing along the way through all the editing processes it may have undergone in intermediate stages. And determining the type and amount of information to be included.

At this point we see a very specific difficulty beyond mere shooting, such as the legitimacy of computer-generated content inserted into our program — illustrations or graphics, for example. In other words, we must be able to identify and show the viewer the provenance and manipulations practically at frame level.

And all this is intended to be achieved through consensus among all stakeholders, who have their own product ranges, services, commercial objectives and technical and business development projects. In this regard, it is worth mentioning and appreciating that virtually all major players are already involved in this initiative.

The purpose of the C2PA consortium is to develop an open technical standard that makes it possible to create and manage content credentials that allow the origin to be identified and all modifications undergone by digital audiovisual content to be tracked

In this sense, we invite you to look closely at the consortium’s website (links at the end) to see what types of companies are highlighted as the 11 steering members, participating with greater involvement and in some way leading the more than 500 already affiliated at the time of writing. We will only offer an interesting clue: there are more service companies than product companies.

Undoubtedly, the project is far more ambitious and complex than it may appear at first glance. Let us therefore move forward by examining the achievements reached, the current situation and the foreseeable future.

To simplify the explanation, we will turn to something we are all already very familiar with: our personal certificates in our Internet browsers. These are certificates issued by a Certification Authority that guarantees their origin and security and that we currently use, for example, to identify ourselves before official bodies or to sign documents.

What is important is that these certificates are not sent when we sign a document with them; rather, the legitimacy of that signature is recorded and can later be verified through various means, such as certification and signature verification platforms. Making it a convenient and secure procedure.

Something similar is what the C2PA coalition seeks to achieve: that there may be all kinds of devices — cameras, editors, distribution channels, players, etc. — capable of supporting and progressing these certificates of provenance and content processing, maintaining that information embedded and immovable within our audiovisual files. The ultimate purpose is that this information can later be verified at any stage of the production chain and even at final viewing.

Just as these certificates are secure in our browsers, they must be equally secure in our equipment, issued by a Certification Authority that makes them reliable and operational across all compatible devices, according to a standard that is already defined and whose functionality has been verified.

But that is not all: in addition to being issued by accredited entities, they must be reflected in every piece of content and every file recorded with that camera and edited with that editor. We will return to this point shortly when discussing the camera and the “real world.”

Returning to our production processes, let us see what happens once the file leaves the camera with the certificate discreetly stamped as a collection of metadata that must coexist with our content and remain unaltered in subsequent processes. For example, once those materials enter our editing process, our certificate must be maintained, and the editing software must recognize and progress it, adding information about the modifications that content has undergone along with its corresponding editor certificate.

The same must occur at every single stage in which the material is modified in any way, until it reaches the final viewer. That is, the broadcaster, as the final party responsible for the content, may — or must — add its own certification.

Of course, we are now all thinking about what kind of image we will see if what appears before us is this enormous amount of overlaid information. Let us remain calm, because initially we will see practically nothing. The idea is that this information may be displayed or not at the viewer’s discretion. In other words, the image will remain clean and will only show a small “cr” logo when validation metadata exists, at which point we may choose whether or not to view the details of those certifications.

We insist that all this has a much broader scope than it may appear at first glance, since it would be very easy for any editor or broadcaster to create an overlay imitating this certification. That is why it is important that content merely indicates whether it contains this information, and that it be the viewer who can activate or deactivate it at will to guarantee the credibility of the procedure. And that all the media involved, such as playback applications, manage it properly.

We are already in a sufficiently advanced phase of development and demonstration to have it accessible and to assess the change it may represent

Although technically the solutions are relatively complex — with the added challenge of achieving agreement among all companies across all sectors beyond their commercial programs — the final objective is that for the user it should be as simple as activating or deactivating subtitles in any player, whether online, local, OTT, etc.

And this requires multiple contributions at all levels. But it is not a utopia. In fact, at the last IBC (Amsterdam, September 2025) it could already be seen in action at different stands. It is true that there are currently very few resources capable of supporting this protocol and these licenses, but what seems truly important and significant to us is that the first ones already exist and they are not prototypes, but commercially available equipment and applications.

Regarding cameras, it seems that for now (verified as of September 2025) there is only one: the new Sony PXW-Z300, which already supports it and has been used in all demonstrations, both inside and outside the IBC exhibition. As for software, the latest version of Adobe Premiere Pro also supports it. Both were demonstrated and driven by the first tests carried out by the British broadcaster BBC.

It is public knowledge that many companies are doing what is necessary for their products and solutions to participate in this initiative. In fact, many are already part of the consortium, at different stages of progress. We dare to predict that it will only be a matter of time before everyone — at least all those involved in news production — has products, services and solutions compatible with this standard.

However, it should also be noted that we are still at a fairly early stage. For example, at this moment the certificate is only embedded and managed in proxy files, but not yet in full-quality MXF files. The reason appears to be related to the need for more elaborate agreements among more participants in order to maintain all the compatibilities that the different formats currently have.

Going even deeper into the details, and following the sequence of a production, let us see what the camera does. First, we must install a certificate that will be linked to the author or owner. Because even at this stage there are different possible actors: a camera may contain the certificate of its owner if it belongs to a freelancer, or the certificate of a production company if it is equipment handled by several operators.

So, is the certificate stamped on all content? Absolutely not. And it is not simply a matter of activating or deactivating it. The camera must use its processing and AI functions to recognize “artificially flat” scenes — such as shooting from a screen — to prevent incorrect stamping and thus help avoid confusion between “the truth of the real world” and “artificial worlds.” Although we do not have all the technical details of what the discrimination threshold would be, it seems that in such cases the certificate would not be stamped, or would be stamped with indications related to this particularity.

We dare to predict that it will only be a matter of time before everyone — at least those involved in news production — has products, services and solutions compatible with this standard

For the next step, we will need an editing system with its own certificate. As with the camera, in September there was only one software solution available: Premiere Pro. What will its role be? Presumably, to maintain validation in those shots where it should be maintained — for example, cropping from a longer shot — adding its own certification when recording the action performed. If we stop at this point to count the number of things that can be done in editing, we would never finish listing examples of “validable” and “non-validable” actions.

As an example, we leave just one question to stimulate debate: is color correction an alteration of reality?

The fact is that after the editing process we will have new content that will contain, at least, the certification of the camera “owner” and those of the different possible “responsible parties” for the edits, which will differ for each fragment (even at frame level) of the finished content. Since that content will be a compilation of images from multiple sources, and not all of them necessarily come from a camera, nor are they necessarily “forgeries” when they do not come from a camera.

And this is where the broadcaster comes into play, who, as planned, may also have its own certificate to stamp — added to all previous ones — with which it shows and records its actions, since depending on the type of content it will have made its own elaborations. Without losing sight of the fact that throughout the entire process we are not only talking about images, but also all the audio associated with them.

To complete the scenario, the broadcaster could display key parts of the veracity or recreated image information as an overlay in the broadcast. In the research and development section of the BBC website there is an interesting report containing a couple of sample videos. We strongly recommend watching the first one, just under three minutes long, paying special attention from minute 1:40 onward. You will also find the link at the end.

With everything we have seen so far — which barely represents a brief outline of what lies behind the initials “C2PA” and “cr” — it is already possible to grasp the magnitude of the initiative they represent, the enormous scope of this project, and the impact it will have in the short term on the way we capture and disseminate credible content.

Because despite the vast number of positive contributions that artificial intelligence offers, what is truly disturbing today is its impact on content credibility. A very powerful tool can be used for multiple purposes. And not being able to distinguish which side of the delicate line separating information from disinformation — and consequently manipulation — we are on would leave us in a worrying situation of vulnerability if we did not have initiatives such as this one.

Finally, so as not to end on a tragic note, we leave you with a new question: will C2PA be the predecessor of some kind of “protocol droid” designed to interact with humans?

After this small joke, here are the links mentioned so that you can explore this initiative first-hand:

• C2PA Consortium: https://c2pa.org
• “Content credentials” watermark or “cr”: https://contentcredentials.org
• BBC report mentioned: https://www.bbc.com/rd/articles/2025-09-news-content-verification-credentials-trust

Despite the vast number of positive contributions that artificial intelligence brings, what is truly disturbing today is its impact on content credibility