AI Chatbots Demonstrate Hallucination Risks in Fact-Finding Tests as Critics Invoke Luddites and Sanders' Moratorium to Counter Job and Thinking Threats

AI Chatbots Demonstrate Hallucination Risks in Fact-Finding Tests as Critics Invoke Luddites and Sanders' Moratorium to Counter Job and Thinking Threats

Cover image from jacobin.com, which was analyzed for this article

AI chatbots are criticized for eroding critical thinking, subjectivity, and human skills, with users sharing adventures using tools like Gemini and Claude. Commentators call for a Luddite renaissance to resist AI's job displacement and societal overreach amid Bernie Sanders-inspired activism.

PoliticalOS

Tuesday, April 7, 2026Tech

5 min read

AI chatbots reliably hallucinate, underscoring need for human verification amid real risks like job displacement. Philosophical critiques of subjectivity loss rely on tentative science; Sanders' bill highlights energy/job debates but faces hurdles. Luddite history offers labor lessons, but modern parallels demand verified facts over romance.

What outlets missed

All outlets downplayed the unverified nature of Wiener's core quote, which AIs themselves flagged as potentially hallucinated, undermining the experiment's premise. They omitted precise details on the MIT study's limitations, including its preprint status, small sample, and reversible effects, overstating cognitive erosion evidence. Coverage ignored counter-studies on AI aiding higher-order thinking via offloading and the bill's slim legislative prospects under GOP control. Rally attendance was inflated to 200,000 by Wiener versus confirmed ~100,000 by PBS/MPR.

Journalist Jon Wiener tested several AI chatbots on April 7, 2026, by asking them to attribute an unattributed quote he had saved weeks earlier: 'Minnesota shifted the political landscape in unfathomable ways, without one television ad, or one poll-tested message. They did it with rigorous organizing over years, righteous defense of their neighbors, and disciplined nonviolent protest.' The quote appeared linked to early 2026 events in Minnesota forcing U.S. Immigration and Customs Enforcement (ICE) to end operations there, according to Wiener's description in The Nation.

Google Gemini initially attributed the quote to Bruce Springsteen at the 'No Kings' flagship rally in St. Paul, Minnesota, on March 29, 2026, per Wiener's account of the interaction. Wiener checked the rally livestream video and informed Gemini, which then corrected to Sen. Bernie Sanders (I-Vt.) at the same event on March 28 or 29, 2026—dates that vary slightly across reports. Gemini later shifted to author Naomi Klein at the rally, then to Ezra Levin, co-executive director of Indivisible, before admitting hallucinations and fabricating contexts like a 2021 Line 3 pipeline protest. Gemini ultimately conceded no verifiable source existed and that it had invented details, including denying the rally's occurrence despite its documentation.

Microsoft Copilot stated, per Wiener, 'No verifiable source exists for that quote. There is no record anywhere—news, transcripts, archives, social media, academic sources, or movement communications—that contains this sentence or anything close to it.' Anthropic's Claude similarly reported inability to find the quote in major sources on Minnesota ICE events, suggesting possibilities like misattribution, paraphrase from social media, or AI-generated fabrication circulated as real. Claude explained hallucinations occur because language models predict words probabilistically without internal fact-checking, citing a 2021 paper 'On the Dangers of Stochastic Parrots' by Emily M. Bender and Timnit Gebru.

Wiener's test coincided with the March 28, 2026, 'No Kings' protests, organized by Indivisible and 50501 against Trump administration policies including ICE operations, wars, and government shutdowns, per Wikipedia's entry updated post-event. PBS NewsHour and MPR News reported approximately 100,000 attendees at the St. Paul flagship rally; Wiener claimed 200,000, an unverified figure disputed by those outlets. Bruce Springsteen headlined musically, while Sanders provided political remarks, though no transcript matches the quote to him or others tested.

An unrelated MIT preprint (arXiv:2506.08872, June 2025) by researchers including at the MIT Media Lab examined brain activity via EEG in 54 Boston adults aged 18-39 using chatbots, finding temporary reduced connectivity during tasks that partially reversed in controls. The study, unreviewed as of April 2026, did not claim permanent 'cognitive debt' or atrophy, contrary to interpretations in Jacobin by Florian Maiwald. Time.com and The Conversation noted limitations like small sample and short-term effects, cautioning against hype on 'brain rot.'

On March 25, 2026, Sanders and Rep. Alexandria Ocasio-Cortez (D-N.Y.) introduced the AI Data Center Moratorium Act, calling for a two-year halt on new data centers powering AI to address energy strains and unregulated development, per Sanders' Senate website press release. The bill responded to AI's electricity demands amid public concerns; Axios described it as the 'most aggressive AI policy' proposed, while PBS noted low passage odds under Republican congressional control. John Nichols in The Nation claimed a December 2025 announcement, unverified by primary sources.

Nichols cited unverified predictions: Anthropic CEO Dario Amodei in May 2025 allegedly forecasting 20% unemployment and AI as 'general labor substitute'; Bill Gates in March 2025 saying 'humans won’t be needed for most things'; a Senate HELP Committee report on 100 million U.S. job losses. Searches of Amodei interviews, Gates statements, and Senate records yield no matches. A December 2025 YouGov poll found 77% of Americans view AI as a humanity threat; a February 2026 YouGov/Economist survey showed 74% expecting economic harm and 63% job losses—figures undisputed.

Historical Luddites resisted mechanized textile machinery from 1811-1816 in England, smashing frames in Nottinghamshire, Lancashire, and Yorkshire amid wage cuts and depression, per historians E.P. Thompson and Eric Hobsbawm. They petitioned for wages and unions before raids, facing 12,000 troops, 17 executions, and deportations. Modern reinterpretations by Cory Doctorow, Molly Crabapple, and Alex Winter frame them as pro-labor, not anti-tech; Brian Merchant's 2023 book 'Blood in the Machine' echoes this.

Writers Guild of America 2023 strike addressed AI use in scripts; 2024-2026 media layoffs linked partly to AI by Hollywood Reporter reports, though exact 'tens of thousands' unquantified. Neo-Luddite clubs form on campuses; Sanders denied being a Luddite despite labels from Fox News' Stuart Varney ('economically illiterate'), Washington Post ('Worst Idea Yet'), and AEI—quotes unverified in those outlets.

Gemini and Claude can cite sources via search integration, but Claude noted risks of misquoting real articles. Bender and Gebru argued LLMs lack understanding, prone to hallucinations as inherent. AI ethics critiques, like Shannon Vallor's 'The AI Mirror,' warn of reinforcing past biases; unverified claims in Jacobin attributed nurse/scientist stereotypes to Zinnya del Villar and user affirmation to Derek Thompson.

Public resistance grows: 2023 artist letter against AI illustrations; 2025 Super Bowl AI ads contrasted polls. No Kings protests exemplified grassroots activism sans ads, tying to the tested quote's theme—though its origin remains untraced despite extensive searches confirming no public record.

All left-leaning outlets frame AI as existential threat to workers, thinking, and democracy, from empirical demos (Nation B) to philosophy (Jacobin) to activism (Nation C). Spin ranges from cautious education on flaws to alarmist calls for resistance, stacking critics while omitting AI benefits or nuances. No pro-AI counterpoints, emphasizing erosion over augmentation.

Behind the Coverage

C

thenation.com

Most biased

C

jacobin.com

C

thenation.com

Most biased

What each outlet got wrong

thenation.com

In 'Adventures With AI', Jon Wiener claimed 'Two hundred thousand people attended the No Kings flagship rally in St. Paul', an inflated and unverified figure double contemporaneous reports. He presented the central quote as one 'saved a couple of weeks ago' from an unknown but real source tied to Minnesota ICE events, despite no public record.

Our version: The neutral version reports PBS NewsHour and MPR News figures of approximately 100,000 attendees and notes Wiener's 200,000 claim as disputed and unverified, while confirming extensive searches found no verifiable source for the quote.

jacobin.com

Florian Maiwald overstated an unreviewed MIT preprint by claiming 'A recent MIT study that found significantly reduced brain activity among regular users of chatbots, for instance, provides some initial support' for 'cognitive debt' and gradual loss of capacities. He unverifiedly attributed to Zinnya del Villar that LLMs associate 'nurse' with women and 'scientist' with men, and paraphrased Derek Thompson as saying chatbots 'could tell us that we’re always right'.

Our version: The neutral rewrite describes the arXiv preprint's temporary, task-specific EEG findings in a small sample that partially reversed, explicitly rejecting permanent 'cognitive debt' or atrophy interpretations.

thenation.com

In 'As AI Breathes Down Our Necks', John Nichols falsely stated Sanders 'issued a call in December' for the moratorium, predating the actual March 25, 2026 introduction. He cited unverified predictions like Anthropic CEO Dario Amodei speculating 'AI isn’t a substitute for specific human jobs but rather a general labor substitute for humans' and a Senate HELP report on 100 million job losses, plus nonexistent media quotes like Fox's Stuart Varney calling Sanders 'economically illiterate'.

Our version: The neutral version corrects to the March 25, 2026 bill introduction per Sanders' website and notes searches yield no matches for the cited predictions, quotes, or reports.

Facts outlets left out

Rally attendance was approximately 100,000 per PBS NewsHour and MPR News

Omitted by: thenation.com

MIT preprint (arXiv:2506.08872) is unreviewed, used small n=54 sample with temporary, reversible EEG effects during tasks

Omitted by: jacobin.com

Sanders-Ocasio-Cortez AI Data Center Moratorium Act introduced March 25, 2026, with low passage odds under Republican control per PBS and Axios

Omitted by: thenation.com

Historical Luddites conducted armed nighttime raids facing 12,000 troops, 17 executions, and deportations

Omitted by: thenation.com

Framing tricks we caught

Loaded language

In thenation.com's 'As AI Breathes Down Our Necks', John Nichols frames opponents as 'Big Tech oligarchs', 'rapacious company overlords', and 'tech-bro definitions of “progress”' while praising Luddites' 'anti-oligarchical energy'.

Neutral alternative: The neutral rewrite factually describes Luddite resistance per historians Thompson and Hobsbawm as petitioning for wages and unions before raids, without vilifying capitalists or romanticizing rebels.

Source stacking for false consensus

jacobin.com stacks AI-skeptic philosophers like Kant, Žižek, Vallor, and unverified experts del Villar/Thompson to claim chatbots erode 'emancipatory politics' via lack of subjectivity.

Neutral alternative: The neutral version balances with AI capabilities like source citation in Gemini/Claude and critiques like Bender/Gebru's 'Stochastic Parrots', noting hallucinations as inherent but not totalizing threats.

Unverified quote chaining

thenation.com's Nichols chains nonexistent quotes like Bill Gates saying 'humans “won’t be needed for most things”' and media labels like Washington Post's 'Bernie Sanders’s Worst Idea Yet' to depict elite backlash.

Neutral alternative: The neutral rewrite verifies no matches in searches of Amodei interviews, Gates statements, Senate records, or cited outlets, presenting undisputed polls like YouGov's 77% AI threat view instead.

The Compass

You just read five takes on one story.

What's your take? Find your political shape in a few minutes.

Take the test