How NOT to Use A.I.: Fake Endorsements from Real People

A well-known (in followers of the tech world, at least) YouTuber and entrepreneur, Marques Brownlee (online-platform handle: mkbdh) accused a company of using gen-A.I. to create a faux endorsement of the company’s product.

The Five Most Key Takeaways from This Blog Post

This is a significant deal because right now lawmakers have still not done much, regulation-wise, to protect people against such deepfakery. (Although an election-protection law covering deepfakes of politicians got through in California.)
A reporter from The Verge describes the deepfake of Marques Brownlee as “imperfect but passable”. That is a good way of describing the potential effectiveness of A.I. deepfakes for fake endorsements, as in the noise of online content, few are looking or listening too closely. As such, it is easy for some companies to get away with this.
A potential mitigation against this practice is watermarking, although that can only go so far. After all, some people may fail to notice the watermark. Or simply fail to remember whether the video was watermarked if the video comes to mind later. (Sort of akin to those “Was that a memory, or did I dream that?” moments.)
The (pretty obvious, really) ethical argument for businesses to not engage in this practice is that it is deceptive. In addition to that, it is baldly mercenary, could damage the reputation of the deepfaked endorser, and tank the company’s credibility among consumers.
For businesses on the straight and narrow, this nefarious use of A.I. could lead to wider skepticism about endorsements in general. That will likely be further augmented by A.I.-generated endorsements that the endorser signs off on in the first place. The question of “is this even real?” could hurt businesses’ chances in endorsement campaigns.

Misendorsement

Obviously, even if most companies are unaware of the particulars of the legal landscape surrounding deepfakes, most would still refrain from such ethically unsound practices.

Endorsements are pretty sensitive territory to begin with, because most of the people doing endorsements have earned some level of trust with a subset of the consumer population.

For marketers, this trust is the inroad to those consumers’ consciousnesses, because the consumers are more likely to listen to, or even just pay attention to, the person doing the endorsement.

But deepfake endorsements further abuse this trust by flaunting the seeking of the endorsers’ approval in the first place.

Most of us recognize the sensitivity of endorsements because of the very practice of seeking the express permission of using a recognizable figure’s name, image, and likeness to promote a product.

That is markedly different from, say, creating a meme featuring that same endorser. The writer of this blog post doubts whether anyone in the history of meme-making has reached out to a recognizable figure for express permission to use the figure in a meme.

Why? Because people’s money, rather than mere attention, is on the line in endorsements. Plus, most people realize that a meme is fake, as opposed to deceptive content that masquerades as authentic.

But of course this has wider implications for people who are not just the deepfaked influencers, or the duped customers who purchase a product based on the deepfake content.

The Real Impact on Business Owners

The widespread erosion of trust in endorsements could result from the continual use of deepfakes for misendorsements.

Another issue is that businesses may have partnerships with endorsers who end up in deepfakes, which could lead to headaches, especially if the deepfake is from a competitor.

Business owners may need to seek more comprehensive ways to authenticate that endorsements are indeed real, and not just deceptive gen-A.I. content.

For instance, maybe having in the contract that the endorser will comment on the endorsement post, signaling the authenticity of the endorsement, could be one method.