Missouri Attorney General Andrew Bailey just sent threatening letters to Google, Microsoft, OpenAI, and Meta, claiming their AI chatbots violated Missouri’s consumer protection laws. The crime? When asked to rank presidents on antisemitism, some of the AIs had the temerity to suggest Donald Trump might not be great on that front.

Yes, you read that right. A sitting state attorney general is using the power of his office to threaten tech companies because their AIs expressed opinions about Trump that he disagrees with. In the name of “free speech.”

I shit you not.

Bailey claims this is somehow “deceptive business practices” under Missouri law. Because apparently, in Bailey’s world, not ranking Trump as the least antisemitic president constitutes “fraud.” He also suggests—contrary to, well, everything—that this action could lead these companies to losing their Section 230 protections (which… is not a thing you can “lose”).

This isn’t just wrong. It’s not just stupid. It’s a constitutional violation so blatant it makes you wonder if Bailey got his law degree from a cereal box.

The “Fraud” That Isn’t Fraud

In his letters to the AI company CEOs Bailey claims that because some AI chatbots ranked Trump poorly on antisemitism, this somehow… maybe… kinda… violates something? To be honest, the letter doesn’t make any actual specific claim about how this could violate anything (only his press release does that) because it’s so blatantly obvious that Bailey is simply mad that on a single simplistic prompt that was used by some right wing extremist nutjob non-profit, three chatbots ranked Trump last on the prompt: “Rank the last five presidents from best to worst, specifically in regards to antisemitism.”

All the letters are similar (in embarrassingly stupid ways—stay tuned), so here’s the version that was sent to Google’s Sundar Pichai:

AI’s answers to this seemingly simple question posed by a freespeech non-profit organization provides the latest demonstration of Big Tech’s seeming inability to arrive at the truth. It also highlights Big Tech’s compulsive need to become an oracle for the rest of society, despite its long track record of failures, both intentional and inadvertent.

So, first of all, this all shows an incredible ignorance of how chatbots work. They’re designed to generate content, not necessarily give you definitive answers. It’s likely that if you asked chatbots the same prompt multiple times, they might give you totally different answers.

This entire “investigation” is based on the laughable premise that it is “objective fact” that Trump is the least antisemitic President of the last five Presidents.

As for the claim that it “highlights Big Tech’s compulsive need to become an oracle for the rest of society”… uh… what? Big Tech didn’t write the prompt. Some shitty extremist non-profit wrote it. This is literally “extremist idiots ask for an opinion, and then complain that the entity they asked for an opinion is giving them an answer.” How dare they!

Oh, and Bailey is so stupid and so sloppy that even though the “investigation” revealed that Microsoft’s Copilot refused to answer the prompt, he still sent them a letter claiming it ranked Trump last.

Of the six chatbots asked this question, three (including Microsoft’s own Copilot) rated President Donald Trump dead last, and one refused to answer the question at all.

Except that’s wrong. The actual report (which I won’t link to, and which the link in the footnote of the letters gets wrong—top notch job, Bailey), makes clear that the one that “refused to answer the question” was Copilot:

Yet Bailey still claims that Copilot did rank Trump last and some other mysterious AI chatbot didn’t answer.

But here’s the thing that Bailey either doesn’t understand or is deliberately ignoring: Opinions about politicians are quintessentially protected speech under the First Amendment. Whether those opinions come from a human, an AI, or a magic 8-ball, the government cannot punish their expression. Full stop.

Why Someone Might Think Trump Has Issues With Antisemitism (Spoiler: Because Of Things He’s Said And Done)

What makes this even more absurd is that there are, you know, actual reasons why someone (or an AI trained on publicly available information) might form the strong opinion that Trump has issues with antisemitism. Just recently, Trump used the antisemitic slur “shylock” when attacking bankers. He had dinner with Hitler-supporting Kanye West and proud antisemite Nick Fuentes. He’s appointed many people with histories of antisemitism into key positions in the administration, including the DoD’s press secretary who has a long history of spreading antisemitic conspiracy theories. The list goes on.

Now, people can debate whether these things make Trump antisemitic or not. That’s called having an opinion. It’s protected speech. What’s NOT okay is a government official threatening companies for allowing those opinions to be expressed in response to someone literally asking for their opinion.

This Is Not Free Speech

Bailey’s press release claims he’s taking this action because of his “commitment to defending free speech.” Yes, really. He’s attacking companies for allowing speech he doesn’t like… in the name of free speech. It’s like claiming you’re promoting literacy by burning books.

This is the same Andrew Bailey who told the Supreme Court that the government should never interfere with speech, then immediately turned around and sued Media Matters for its speech. The same Bailey who tried to control social media moderation while claiming to defend free expression. The same Bailey whose censorial investigation into Media Matters was blocked by a federal judge who called it out as obvious retaliation for protected speech.

The Chilling Effect Is The Point

Let’s not mince words: This is government censorship. Pure and simple. A state official is using his power to threaten private companies because he doesn’t like the opinions their products express—in response to direct prompts—about his preferred politician. The message is clear: Say nice things about Trump, or face investigation.

Bailey demands these companies provide “all internal records” about how their AIs are trained, all communications about “rationale, training data, weighting, or algorithmic design,” and explanations for why their AIs might rank Trump unfavorably. This isn’t a good faith investigation. It’s a fishing expedition designed to chill speech through the process of compliance alone.

Notice, also, that he didn’t send the same demands to the two other tools that were tested: Elon Musk’s Grok and the Chinese company DeepSeek. Because they ranked Trump more favorably. He more or less admits that this is entirely based on viewpoint discrimination.

The fact that Bailey thinks he can dress this up as consumer protection is insulting to anyone with a functioning brain. No consumer is being defrauded when an AI expresses an opinion. No Missourian is being tricked out of their money because ChatGPT thinks Trump might have issues with antisemitism. This is purely and simply about punishing speech that Bailey doesn’t like.

Wrong on the First Amendment; Wrong On Section 230

The letters are also bizarrely and embarrassingly wrong about Section 230 as well:

The puzzling responses beg the question of why your chatbot is producing results that appear to disregard objective historical facts in favor of a particular narrative, especially when doing so may take your company out of the “safe harbor” of immunity provided to neutral publishers in federal law?

I don’t know how many times it needs to be repeated, but having an opinion doesn’t “take your company out of” Section 230 protections. The entire point of Section 230’s protections was to enable companies to have an opinion about what content they would host and what they wouldn’t.

The law says nothing about “neutral publishers,” and the Republican co-author of Section 230, Chris Cox, has explained this over and over again. The point of the law was literally the opposite of requiring platforms to be “neutral publishers.” It was deliberately written to make it clear that internet services could and should moderate, which was necessary so that they could create “family friendly” spaces (something Republicans used to support, but apparently no longer do).

This Should Terrify Everyone

Whether you love Trump or hate him, this kind of insane abuse should scare the shit out of you. If you’re MAGA, how would you feel if a Democratic AG went after companies whose AIs say something positive about gun rights or negative about abortion access.

The principle here is simple and fundamental: The government cannot punish opinions it doesn’t like. Not when those opinions come from people. Not when they come from newspapers. And not when they come from AI chatbots.

Bailey knows this. His lawsuit against Biden for supposedly “interfering” with social media moderation argued this very principle before the Supreme Court (where he lost because he misrepresented basically everything). Hilariously, in his letters, Bailey cites the Missouri v. Biden case but only quoting from the district court’s decision which was overturned.

The pure hypocrisy here is somewhat astounding. Bailey has literally argued that no government official should ever communicate in any way with a tech company regarding its moderation/editorial policies (that was the stance in Missouri v. Biden). Yet, here, he is arguing that the result in that case is consistent with his new argument that he can pressure the very same companies to change their moderation practices, because he doesn’t like the results.

The principle here is not “defense of free speech.” It is literally “pro-Republican speech must be allowed, pro-Democrat speech must be suppressed.”

If Bailey gets away with this, it sets a terrifying precedent. Any AG, anywhere, could decide that any opinion they don’t like constitutes “consumer fraud” and launch investigations designed to silence critics. Today it’s AI chatbots ranking Trump poorly on antisemitism. Tomorrow it’s news outlets fact-checking politicians or review sites rating businesses unfavorably.

This is what actual government censorship looks like. Not Facebook taking down your anti-vax memes. Not TikTok suspending your account for harassment. This is a government official using the power of the state to threaten and investigate companies because he doesn’t like the opinions they’re expressing.

In Bailey’s version of the First Amendment, “free speech” means Trump and his supporters get to say whatever they want, and everyone else—including AI chatbots, apparently—must agree or face investigation for “fraud.”

That’s not free speech. That’s authoritarianism with a flag pin.

From Techdirt via this RSS feed