Shiffrin Versus the AI Slopocalypse

Shifrin develops a rather thorough account of the constitutional obligations of government officials to be truthful and to uphold the protections of the press. She very categorically writes that “lies and culpable misrepresentations by government officials about public affairs violate the Free Speech Clause, whether or not those misrepresentations are believed or cause harm to particular individuals or groups” (993). While abstaining from taking an opinion on this account, I would like to pose the question of those potential "culpable misrepresentations" currently being propounded by the government, particularly the White House's recent dissemination of AI-generated propaganda media, or “AI Slop” in colloquial terms. 

Shifrin defines culpable misrepresentations as “false statements on a topic about which the speaker has a special responsibility to speak accurately, but culpably derelicts his duty” (1102). Just in the past months, official government accounts have posted images depicting President Trump as a messianic figure, or cartoonish renditions of crying deportees, even Trump or engaged in valiant acts of dominance (defecating in a plane) over political opponents (over a “No-Kings” Protest). These images, posted through official government media accounts, often Trump’s own, represent a novel category of government speech.  They are certainly “culpable misrepresentations” of reality, yet they differ from the definition discussed by Shiffrin in that they maintain a level of plausible deniability for those who posted them. Unlike a verbal misrepresentation that can be categorized semantically, like the false claims about election fraud or pandemics, an AI-generated image does not assert a proposition in any straightforward sense, and cannot be deemed an insincere misrepresentation. It does not claim “X is true and Y is false” in a way that could be evaluated against what the poster believes, or can be corrected by a fact-checker, or prosecuted under a standard of “culpable misrepresentation”. The Trump-as-Jesus image does not explicitly say Trump is divine and the plane defecation video does not explicitly claim No-Kings protestors are an enemy of the state.

No one in the White House communications office sincerely believes Trump defecated on No-Kings protestors from an airplane, and no one sincerely believes that he descended down in white robes and a heavenly aura. Yet at the same time, it is not even a lie in the classical sense, AI Slop is merely satirical. So, how should one adjudicate if governments should be allowed to produce and publish such media?

Shiffrin's argument against government lies is advanced by a diagnosis of what makes them constitutionally defective; that they corrupt the medium through which nations communicate with their denizens, and vice versa. As she writes, "lies introduce insincerity into the testimonial channel, the function of which depends upon its exclusive use as a medium for conveying sincere beliefs" (1011). For Shiffrin, the testimonial channel is the communicative infrastructure of democracy, the public broadcast of public affairs; a true pillar of democratic governance. Thus, when government officials lie or add insincere “noise” into the testimonial channel, they degrade the communicative channel itself, which “introduces high additional filtering costs to identify sincere speech” and insidiously “weakens the morale of audiences to invest in communicative exchanges”, subverting the literal definition of democratic participation. (1010, 1011).

This again complicates the question of whether governments should be precluded from posting AI-generated media. I am of the opinion that AI slop can be categorized as noise because it is volumetrically overwhelming in a way that no individual statement could be, yet it amounts to nothing in the real world. Shiffrin correctly worries about filtering costs “rationally diminishing confidence”  in public channels, but does this filtering cost not already exist without the government? Modern media is dominated by users posting millions of AI-generated images and videos online, and ordinary users already have to distinguish what is real and what is fake. Therefore, whilst the government is certainly weakening the democratic channel, I believe the government’s burden of blame for creating “noise” is paltry, and they are merely an involved party in an ostentatious crowd of AI-generated voices. 

Shiffrin's prescription for the enforcement problem is, ultimately, a turn toward private actors, in this case social media platforms. A platform that removes Trump-as-Jesus must articulate why a cartoon crosses a constitutional line, and must do this without the existence of established corporate doctrine. Yet Trump (as a private citizen) is the majority shareholder of Truth Social, where he posts much of his AI-generated media propaganda, so it is safe to say that government-sponsored AI Slop will be here to perturb our online lives for the near future.