“Content Provenance” in Artificial Intelligence: What Business Owners Should Know

The Five Most Key Takeaways from This Blog Post

There have been many efforts to mitigate the problem of identifying A.I.-generated content on the web. One of the more prominent solutions that has plenty of industry-wide support is “content provenance”, which provides A.I.-generated content with cryptographic identifiers that can make it readily identifiable as either A.I.-generated or “authentic”.
The number of motivations to this are numerous. For one, it can counter misinformation, disinformation, and chaos-sowing deepfakes that bad actors use A.I. to proliferate across online media environments. It can also, on the other hand, counter the growing skepticism toward what is seen online by providing an authenticating process.
That being said, there are also plenty of challenges as well. One of the biggest challenges is obtaining wide cooperation and collaboration between more than just A.I. companies, but a wide range of companies and entities providing relevant software (such as image-hosting platforms) and hardware (think of cameras). This would be crucial for ensuring interoperability across the many photo-generating and -hosting technologies in the world.
One of the foremost forces in this effort is the Coalition for Content Provenance and Authenticity (C2PA) and the Content Authenticity Initiative (CAI, which is easy to misread as the initials of a certain American agency that gathers intelligence that is central to American safety and interests).
For business owners, advancements in content provenance could mean that customers, potential and current, could be readily alerted to any use of A.I. in marketing.

Content Provenance and Business Owners

That last bullet point does make the ears perk up a bit, doesn’t it?

Because imagine for instance if your business puts out a marketing video that used A.I. for some flair.

Now, not everything in the video is A.I.-generated. This content piece is something like a collaboration between the business’ internal marketing team and an A.I. platform (take your pick of which platform).

The trouble here is that part of the video features authentic (=not A.I-generated) testimony from someone who loves the company’s product. Real, authentic testimony that viewers may doubt the authenticity of because an authenticator correctly adjudged the whole video as having some A.I. fingerprints on it.

This, then, is the risk that business owners will need to be aware of in using A.I. for marketing efforts. That is, that people may doubt that anything in the video is real because it got flagged by a content-provenance platform.

Given that possibly upcoming challenge in online marketing, it is worth considering the question that is the header of the section below.

How Likely Will Content Provenance Become Commonplace?

The considered opinion of this writer is that most of this will depend on how aggressively politicians pursue ensuring content-provenance measures become law.

Otherwise, you will likely see public pledges from tech companies that by their own accord may not solve the interoperability challenge.

But there is indeed plenty of motivation for politicians to pursue such regulations that mandate content provenance. After all, one of the most decidedly 21st-century problems that politicians now face are deepfakes and accusations from opponents of using A.I. deepfakery to influence voters.

So, considering that politicians’ potential to win elections can weaken due to deepfakery (or grow by unethically embracing its deceptive potential), it is likely that politicians will out of the spirit of self-preservation make serious efforts to mandate A.I. authentication for the future.

The Final Key Takeaway

Whether content provenance does as much good as it promises is besides the point, because the benefits are so obvious. Of course, it cannot solve every problem, such as the possibility that skepticism will remain for a significant number of people who merely adjudge the authentication process as being as doubtable as the content in question.