Elon Musk has said Grok, the A.I.-powered chatbot that his company developed, should be “politically neutral” and “maximally truth-seeking.”
But in practice, Mr. Musk and his artificial intelligence company, xAI, have tweaked the chatbot to make its answers more conservative on many issues, according to an analysis of thousands of its responses by The New York Times. The shifts appear, in some cases, to reflect Mr. Musk’s political priorities.
Grok is similar to tools like ChatGPT, but it also lives on X, giving the social network’s users the opportunity to ask it questions by tagging it in posts.
One user on X asked Grok in July to identify the “biggest threat to Western civilization.” It responded that the greatest threat was “misinformation and disinformation.”
“Sorry for this idiotic response,” Mr. Musk groused on X after someone flagged Grok’s answer. “Will fix in the morning,” he said.
The next day, Mr. Musk published a new version of Grok that responded that the greatest threat was low “fertility rates” — an idea popular among conservative natalists that has transfixed Mr. Musk for years and something he has said motivated him to father at least 11 children….
Category Archives: cheap speech
“Newsmax pays $67 million to settle defamation case linked to 2020 election coverage”
NPR:
Newsmax will pay $67 million to settle one of the last outstanding defamation lawsuits against a news organization for airing false claims that the 2020 election was rigged.
Denver-based Dominion Voting Systems – the same voting-technology company that had received a $787 million settlement from Fox News over its election coverage – brought the lawsuit against Newsmax. A trial was scheduled to begin in October.
In the lawsuit, filed in the months after the 2020 election, Dominion accused the cable news network of spreading false claims that the company’s voting technology had been manipulated to help Joe Biden beat Donald Trump. Like other right-wing news networks, Newsmax featured Trump allies who promoted these conspiracies, including former Trump campaign lawyer Sidney Powell and supporter Mike Lindell of My Pillow.
Newsmax announced the settlement in an Aug. 15 filing with the U.S. Securities and Exchange Commission. According to the document, the network paid $27 million of the settlement on that day; the rest will be paid by January 2027.
Multiple court rulings and investigations by election officials have found no widespread fraud was present in the 2020 election; even still, these debunked claims were still being echoed by factions of Trump supporters in 2024. Dominion has said the election lies caused the company and its employees extensive harm, including death threats and lost revenue….
“Russia is quietly churning out fake content posing as US news”
A pro-Russian propaganda group is taking advantage of high-profile news events to spread disinformation, and it’s spoofing reputable organizations — including news outlets, nonprofits and government agencies — to do so.
According to misinformation tracker NewsGuard, the campaign — which has been tracked by Microsoft’s Threat Analysis Center as Storm-1679 since at least 2022 — takes advantage of high-profile events to pump out fabricated content from various publications, including ABC News, BBC and most recently POLITICO.
This year, the group has focused on flooding the internet with fake content surrounding the German SNAP elections and the upcoming Moldovan parliamentary vote. The campaign also sought to plant false narratives around the war in Ukraine ahead of President Donald Trump’s meeting with Russian President Vladimir Putin on Friday.
McKenzie Sadeghi, AI and foreign influence editor at NewsGuard, said in an interview that since early 2024, the group has been publishing “pro-Kremlin content en masse in the form of videos” mimicking these organizations.’“If even just one or a few of their fake videos go viral per year, that makes all of the other videos worth it,” she said.
While online Russian influence operations have existed for many years, security experts say artificial intelligence is making it harder for people to discern what’s real.
Storm-1679 developed a distinct technique in 2024 for combining videos with AI-generated audio impersonations of celebrity and expert voices, according to Microsoft’s Threat Analysis Center….
Free Access to New Cambridge U Press Book: “The Future of Press Freedom: Democracy, Law, and the News in Changing Times”
This is a great collection edited by RonNell Andersen Jones and Sonja West, and even better that it is open access. (Full book pdf.) I’m honored to have the final version of my chapter here, “From Bloggers in Pajamas to the Gateway Pundit: How Government Entities Should and Do Identify Professional Journalists for Access and Protection.”
Yale “MFIA Clinic Report Provides Roadmap for Attorneys to Challenge Election Disinformation”
Important report released by Yale Media Freedom and Information Access Clinic:
Students from Yale Law School’s Media Freedom and Information Access (MFIA) Clinic have released a new white paper titled “Using the Ku Klux Klan Act to Combat Election Disinformation: A Guide for Practitioners.”
The guide offers a roadmap for attorneys seeking civil remedies against certain forms of digital election disinformation, such as lies about how to vote, impersonation of candidates or officials, and misinformation intended to intimidate voters.
Read the white paper: Using the Ku Klux Klan Act to Combat Election Disinformation: A Guide for Practitioners
The report focuses on two underutilized provisions of the Ku Klux Klan Act of 1871: the “Support-or-Advocacy” clauses of 42 U.S.C. § 1985(3) and the companion “Neglect-to-Prevent” provision of 42 U.S.C. § 1986. Though originally passed to combat Reconstruction-era voter intimidation by the Ku Klux Klan, these statutes remain powerful tools for modern election protection, according to the clinic. The white paper argues that § 1985(3) can provide a private cause of action when election-related disinformation arises from a conspiracy and amounts to a common-law tort (such as intentional interference with the right to vote, misappropriation of likeness, or false-light invasion of privacy) carried out “on account of” someone’s support for a federal candidate.
Additionally, § 1986 may create liability for third parties like robocall vendors, group leaders, or public officials who knowingly fail to prevent such conspiracies when they have the power to do so.
While the First Amendment rightly shields much false speech about elections, the report outlines scenarios where challenges to election disinformation may prevail despite the First Amendment, such as when the disinformation at issue constitutes a lie about election mechanics or an impersonation of a candidate, or when it is particularly likely to undermine election integrity. Drawing on legal precedent, historical context, and real-world examples — including social media ads that promote “texting to vote” and robocalls that impersonate candidates — the report offers a path forward for legal practitioners aiming to challenge harmful election lies without infringing on protected speech….
(Disclosure: I gave feedback on an earlier version of this report as well as worked with the clinic and Protect Democracy on the amicus brief filed in the Mackey case.)
“The Fact Checker rose in an era of false claims. Falsehoods are now winning.”
Glenn Kessler’s final WaPo column as their fact checker:
In reviewing many of the some 3,000 fact checks I have written or edited, there is a clear dividing line: June 2015, the month Donald Trump rode down the Trump Tower escalator and announced he was running for president….
In ending its work with fact-checkers, Meta chief executive Mark Zuckerberg falsely claimed that fact-checkers censored free speech by being “too politically biased,” echoing Trump administration arguments. The Washington Post did not participate in the Meta program, but any Facebook user had the option to opt out of having posts fact-checked. Many fact-checkers would liken their work to nutritional labels on snack foods — providing more information about online content. People are free to ignore the warnings, just as people can ignore nutritional labels….
In 2016, Trump’s opponents still cared about the facts. Florida Gov. Jeb Bush’s (R) campaign had a wall where they posted positive fact checks. Ohio Gov. John Kasich (R) dropped a talking point simply in response to my question for a possible fact check. Hillary Clinton’s staff worked hard to find policy experts to vouch for her statistics. (Her comments on her private email server were less defensible).
But Trump didn’t care. He kept rising in the polls and eventually won the presidency. Other politicians took notice and followed his lead.
Besides Trump, something else changed the nature of truth in the mid-2010s: the rise of social media. The Fact Checker was launched in 2007, one year after the creation of Twitter and when Facebook had only 50 million users. By 2012, Facebook had 1 billion followers; it reached nearly 1.6 billion in 2015. Trump adroitly used Twitter — where he had 2.76 million followers at the start of 2015 — and other social media to spread his message. Trump’s call to ban Muslims from entering the United States was the most talked about moment on Facebook among the 2016 candidates in all of 2015, according to Facebook data.
Social media helped fuel the rise of Trump — and made it easier for false claims to circulate. Russian operatives in 2016 used fake accounts on social media to spread disinformation and create divisive content — tactics that led companies such as Meta to begin to use fact-checkers to identify misleading content. But the political forces which benefited from false information — such as Trump and his allies — led a backlash against such efforts, saying it was a form of censorship. Now tech companies are scaling back their efforts to combat misinformation….
During Trump’s first term, The Fact Checker team documented that he made more than 30,000 false or misleading claims. Week after week, I would write fact checks unpacking his latest misstatements, and Trump generally earned Four Pinocchios — the rating for a whopper. But I sense that the country has gotten so used to Trump exaggerating the truth that it no longer seems surprising. I chose not to repeat the exercise in his second term…
“Project Veritas Withdraws Lawsuit Against The New York Times”
NYT:
The conservative group Project Veritas this week dropped its yearslong libel lawsuit against The New York Times.
The lawsuit accused The Times of defamation for an article published in 2020 that reported that researchers from Stanford University and the University of Washington had described some videos produced by Project Veritas as probably part of a coordinated disinformation effort. The group also sued the researchers.
Project Veritas lost its defamation claims against the university researchers in 2022, and was ordered to pay Stanford nearly $150,000 in legal fees. But the group had continued to pursue its claims against The Times after defeating the news organization’s motion to dismiss.
“We are pleased that Project Veritas decided to withdraw its libel suit without any settlement,” Charlie Stadtlander, a Times spokesman, said in a statement, adding, “The claim against The Times should never have been brought.”
Rove on the Political Price of Conspiracy Theories
Karl Rove in WSJ, connecting the ongoing Jeffrey Epstein drama to election-related conspiracy theories:
Other conspiracies await action by the White House….
Team Trump could produce evidence that confirms the conspiracy advocates were right. They could reveal just how the 2020 election was stolen—computers in Europe, miscounted ballots, phony election returns, etc.—and identify the Deep State agents who conceived this crime against our democracy and bring them to justice. Team Trump could also produce the evidence that Jan. 6 was a Deep State plot to discredit the president by causing law-abiding Americans to breach police lines and storm the Capitol.
But in truth there is no “there” in either case. Neither happened. Years have passed, yet there has been no real evidence or successful court case proving the vast conspiracies around the 2020 election or Jan 6.
However, if Team Trump admits the election wasn’t stolen and government agents didn’t organize the Capitol assault, MAGA conspiracy theorists would be furious and the president’s base further weakened and fractured.
Second Circuit Unanimously Reverses Conviction in Mackey (“Ricky Vaughn”) Case, Finding Insufficient Evidence of a Conspiracy to Deprive People of Their Right to Vote By Tricking Them into Voting By Text
You can find the court’s opinion at this link. The sole basis for the reversal was lack of sufficient evidence that defendant Mackey had conspired with others to deprive people of their right to vote.
The Court did not address the substantive scope of section 241 or address the First Amendment issues with convicting people of depriving others of the right to vote by fraud. I filed an amicus brief with Protect Democracy and the Yale Media Freedom and Information Access Clinic on these issues.
I explained those issues in this post:
and in this piece at Slate.
So resolution of these constitutional and statutory issues will await another day.
“A.I. Is Starting to Wear Down Democracy”
That’s the headline in the NYT, and the second exaggerated NYT headline I’ve flagged today. I don’t know that the article actually delivers what the headline promises (and headlines are usually not, as I understand it, written by the reporters themselves, which continues to strike me as a disservice to the reporters in situations like these).
But if you skip the headline and read the piece itself, you’ll find a really useful catalog of some of the ways in which AI has been used in recent elections, with embedded multimedia examples.
“Voters beware: 25 states restrict AI in elections. SC is in the other half.”
As the Daily Gazette points out, though 25 states have either bans or rules about disclosures for AI in campaign materials, the current draft of the “Big Beautiful Bill” has a 10-year moratorium on state AI regulations that would likely block enforcement of many of these statutes.
FTC ad placement consent decree with a … very unusual provision
The FTC announced its approval today of a consent agreement governing the acquisition by Omnicrom of The Interpublic Group – both offer (inter alia) media buying services. The agreement allows the purchase but prohibits Omnicrom from directing ad spending to or away from any publisher based on the “political or ideological viewpoints” of the publisher or of nearby content on the publisher’s platform unless it’s pursuant to the terms of a specific agreement between Omnicrom and the company doing the advertising and purchasing Omnicrom’s services. And even if one company has an explicit banned- or preferred-publisher list (or asks Omnicrom to come up with one), the agreement stops Omnicrom from sharing that list with any other client.
The agreement sure seems like it’s going to make it significantly harder for Omnicrom to effectuate a client’s “Please don’t show my ad next to Nazi propaganda” wish. But also, on its face, it also seems like it’s going to make it significantly harder for Omnicrom to create a business model favoring certain kinds of “political or ideological” advertising, or placing ads targeting Republicans or Democrats?
Unless Omnicrom is effectively a common carrier – and I’ll confess I don’t know the shape of the industry enough to know – it’s hard to see how this squares with the company’s First Amendment rights…
“Republicans are flagged more often than Democrats for sharing misinformation on X’s Community Notes”
Thomas Renault , Mohsen Mosleh, and David G. Rand in PNAS:
We use crowd-sourced assessments from X’s Community Notes program to examine whether there are partisan differences in the sharing of misleading information. Unlike previous studies, misleadingness here is determined by agreement across a diverse community of platform users, rather than by fact-checkers. We find that 2.3 times more posts by Republicans are flagged as misleading compared to posts by Democrats. These results are not base rate artifacts, as we find no meaningful overrepresentation of Republicans among X users. Our findings provide strong evidence of a partisan asymmetry in misinformation sharing which cannot be attributed to political bias on the part of raters, and indicate that Republicans will be sanctioned more than Democrats even if platforms transition from professional fact-checking to Community Notes.
“Musk’s X Sues New York Over Law Requiring Hate-Speech Data”
X challenges the constitutionality of New York law requiring social media companies to disclose how they deal with hate speech, extremism, and disinformation. Similar to the California law struck down by the Ninth Circuit, “[t]he law, which goes into effect this week, requires X and other major social media companies to file reports detailing how they define and moderate hate speech, racism, extremism, radicalization, disinformation and misinformation, harassment and foreign political interference. It’s part of a broader effort by states to mitigate what they see as the harmful effects of online platforms.”