Category: Fact Check

A recent Facebook post urged readers to search online to find out more information about an image that supposedly showed a cavernous, multi-storied world under the U.S. Capitol building.

“PREPARING FOR SOMETHING,” the March 11 post said. “Google it.”

The image appears to show eight floors beneath the Capitol, including rotundas, ornate architecture and multiple levels of well-trafficked roads.

“Underneath the Capitol building,” text above the image says. 

This post was flagged as part of Meta’s efforts to combat false news and misinformation on its News Feed. (Read more about our partnership with Meta, which owns Facebook and Instagram.)

(Screenshot from Facebook)

A small watermark in the image’s corner led us to synthographyart.com, a website describing itself as a “robust collection of AI-generated images” that represent “a cutting-edge form of artistic expression” that uses synthetic and photographic elements.

The U.S. Capitol building covers 1.5 million square feet with more than 600 rooms and miles of corridors, but we found no evidence that this image reflects that. 

There’s a Capitol subway system with three lines, two on the side with the Senate chambers and one on the side with the House chambers. 

And, according to Atlas Obscura, “a sprawling pedestrian tunnel system under Capitol Hill allows staffers and members of Congress to securely traverse the office buildings, Library of Congress and Capitol building largely out of sight.”

The online “D.C. Underground Atlas,” which Bloomberg News described as a “highly detailed interactive history of how underground D.C. came to be,” discusses the tunnels under the Capitol and abandoned features of the subterranean structures, such as a “Studebaker automobiles” that were later replaced with monorails. 

In the early 2000s, according to the site, “excavators shoveled out a massive-three story pit under the east plaza” that more than doubled the Capitol footprint” for the Capitol Visitor Center, which is entirely underground.

We contacted the Architect of the Capitol’s office for comment but didn’t hear back. However, we found nothing to support the idea of an underground akin to the one in the 1927 movie “Metropolis.”

We rate claims this image shows the Capitol underground False. 

 

When President Joe Biden was in Milwaukee on March 13, Brian Schimming, chairman of the Republican Party of Wisconsin, issued a statement saying:

“On Joe Biden’s watch, Wisconsin has lost 6,000 manufacturing jobs and 455 dairy farms in the last year. Instead of paying us a visit to brag about his abysmal record, the President should be offering working families an apology for Bidenomics.”

That sounds like a lot. Are those numbers accurate? 

Has Wisconsin lost 6,000 manufacturing jobs in a year?

Schimming’s statement on the manufacturing jobs included a link to a graph from the St. Louis Federal Reserve that shows the number of employees in manufacturing over time. 

When we clicked on the graph and went back to February 2023, as the statement claimed, it showed Wisconsin having roughly 481,600 manufacturing workers. When you look at the most recent data, January 2024, it showed 479,300 people working in manufacturing.

So according to the St. Louis Fed, Wisconsin lost 2,300 manufacturing jobs. Far from the 6,000 that was claimed. 

When we asked the Republican Party of Wisconsin about this, it said part of the statement was wrong and the timeline should have been September 2022 to September 2023. 

(In September 2022, Wisconsin had roughly 484,300 manufacturing workers and a year later the state had 477,000 workers in manufacturing. During that time Wisconsin lost 7,300 manufacturing jobs.)

Beyond that, the decline in manufacturing is a long trend fueled by many forces.

The St. Louis Fed has data on manufacturing jobs in Wisconsin going back to 1990. In that time, manufacturing jobs in Wisconsin were at its peak in March 2000 with 600,000 manufacturing employees. 

Recessions in the early 2000s and 2008 had a major impact on manufacturing. In January 2010, manufacturing jobs were at their lowest in the state with 426,200 workers. 

The loss of manufacturing jobs has lasted through multiple presidential administrations from both parties. 

But since 2010, manufacturing has rebounded to 484,900 in June 2019. In 2020, just before the coronavirus pandemic struck, manufacturing jobs were at 477,300 in March. A month later, the number was down to 439,700 jobs. 

In August 2022, manufacturing jobs were at 484,300, completely recovered from pre-pandemic times. Even the current number of 477,000 jobs is near where the industry was before the pandemic. 

So, Skimming was wrong on the numbers and wrong to blame Biden exclusively.

What about the dairy farms?

The Republican Party of Wisconsin provided a link to the Dairy Producer, which cited a report from the Wisconsin Department of Agriculture, Trade and Consumer Protection that said the state lost 455 dairy farms in the last year. 

The department confirmed the farm losses are correct. 

On Jan. 1, 2023, the state had 6,116 dairy farms and on Jan. 1, 2024, the state had 5,661 dairy farms or 455 fewer.

So the state GOP is right on that number, but like the manufacturing jobs, the number of dairy farms in Wisconsin has been declining for decades through multiple presidential administrations.

According to the U.S. Department of Agriculture National Agricultural Statistics Service, in 1992, Wisconsin had 30,156 dairy farms. 

Multiple factors have affected dairy farms. Foreign markets have created global competition for dairy products, the industry depends on immigrant labor while the country wrestles with border policies, and farmers are faced with a choice of getting bigger or losing their farms.

However the number of dairy cows in the state is about 1.27 million and has been at or near that number since 2001. Essentially meaning the smaller farms are selling their cows to larger farms and going out of business while dairy production remains fairly steady.. 

In a 2019 visit to Wisconsin, former President Donald Trump’s agriculture secretary, Sonny Perdue, told reporters: 

“Now what we see, obviously, is economies of scale having happened in America — big get bigger and small go out. … It’s very difficult on economies of scale with the capital needs and all the environmental regulations and everything else today to survive milking 40, 50, 60 or even 100 cows, and that’s what we’ve seen.”

Our ruling

Schimming criticized Biden on manufacturing jobs and dairy industry, saying:

“On Joe Biden’s watch, Wisconsin has lost 6,000 manufacturing jobs and 455 dairy farms in the last year. Instead of paying us a visit to brag about his abysmal record, the President should be offering working families an apology for Bidenomics.”

On manufacturing, he is wrong. Wisconsin lost manufacturing jobs but not near the amount in the claim. On the dairy farms, he is correct. 

But on both points, he is mistaken in placing the blame solely on Biden – manufacturing jobs and dairy farms have been trending downward for years through multiple presidential administrations from both parties. And, the daily industry — when measured differently — still appears strong in Wisconsin. 

So we rate this claim Mostly False. 

   

      

 

A bizarre claim circulating on social media says that actor Mel Gibson released a video of former President Bill Clinton and former Secretary of State Hillary Clinton eating human flesh with a Haitian gang leader named “Barbecue.” 

But no such video exists on the internet. 

An X post has a collage of images that include Mel Gibson speaking into a microphone, Bill Clinton near a car, people riding in a trolley, three men leaning against a white wall while holding rifles and a woman opening her mouth while crossing the street. 

The image is accompanied with text that says, “BOMBSHELL: Mel Gibson just released a VIDEO of Bill and Hillary Clinton engaging in C@NNIBALISM with the Haitian C@nnibal Gang Leader named ‘Barbecue!’”  

This post was flagged as part of Meta’s efforts to combat false news and misinformation on its News Feed. (Read more about our partnership with Meta, which owns Facebook and Instagram.)

The post has a link that redirects users to a news article by American Media Group, a website that shares conspiracy theories. The news article does not contain the video or have any evidence that the video exists.

Alan Nierob, Gibson’s publicist, told PolitiFact that Gibson did not release such a video.

The post’s photo of Gibson is from a 2022 film, “On the Line.” The photo of Bill Clinton and the trolley is from 2010 in Bangkok, when he visited the set of “The Hangover Part II.” 

And the image of the man holding the gun is Jimmy Chérizier, also known as Barbecue, the leader of Haiti’s powerful G9 Family and Allies gang. 

Some people on social media have suggested that Barbecue’s name is a reference to cannibalism, although there is no evidence he or any Haitian gangs are engaging in the practice. Chérizier told The Associated Press in 2019 that he was given the nickname because his mother sold fried chicken when he was growing up. He was sanctioned by the United Nations in 2022 for alleged human rights abuses. 

We searched news reports online and did not find any connection between the Clintons and Chérizier. The Clinton Foundation did raise $30 million for aid projects in Haiti after the 2010 earthquake, with critics saying the money was used improperly. 

We rate the claim that Mel Gibson released a video of Bill and Hillary Clinton engaging in cannibalism with Haitian cannibal leader ‘Barbecue’ Pants on Fire! 

RELATED: Cannibalism in Haiti? Fact-checking the unfounded claims 

An Instagram video opens with a dramatic scene: A man hangs onto a pole as floodwaters swirl around him. Flash floods deluge roads and towns.

Text at the top of the March 10 video says, “Saudi Arabia, (United Arab Emirates), March 9, 2024.” Text at the bottom says, “Climate is changing due to 12,000 years cycle.” Later in the footage, text appears on screen that says, “To create a public demand for solving climate problems, share this post, tell your friends and acquaintances about climate change on your social networks.”

The Instagram video was flagged as part of Meta’s efforts to combat false news and misinformation on its News Feed. (Read more about our partnership with Meta, which owns Facebook and Instagram.)

But the video does not show flooding in Saudi Arabia or the United Arab Emirates. Some of the footage is from Oman, a different country in the Persian Gulf region that has had severe flooding recently.

A March 9 Instagram video on an account called Oman Weather Updates showed the man hanging onto the pole in the March 10 Instagram video and described him as being stuck “in a wadi (a valley) near Madhah Dam and he was rescued by Royal Oman Police.”

From the two- to nine-second mark in the March 10 video, a flooded market street is shown. The names of two shops — Anwar al Khaleej Markets and Taiba Modern Bakery — are visible in the footage. A Google Maps search shows the neighboring shops are located on the al Jmea Road in Al Buraimi, Oman.

A Feb. 12 photo on X also shows flooding in front of Anwar al Khaleej Markets.

While the UAE experienced intense rainfall Mar. 9,  we found no news reports of flooding in Saudi Arabia that day.

Similar claims have said footage showed flooding in one Arab country when it was another. After last September’s floods in Libya that killed more than 11,000 people, social media posts misleadingly shared videos of flooding in Saudi Arabia that occurred seven years before.

We rate the claim that this video shows flooding in Saudi Arabia and the United Arab Emirates on March 9, 2024, False.

When the moon’s shadow briefly darkens cities across the U.S. on April 8, it won’t deplete fuel and food supplies or knock out cellphone service. 

Before the total solar eclipse, officials in states and towns along the eclipse’s path of totality are encouraging residents to prepare for large crowds; some social media users have shared the guidance without critical context.

Emergency officials warn to “stock up on food, water, and fuel ahead of the eclipse,” read the text in a March 14 Facebook reel. “Cell phone service could be a problem too.” 

This post was flagged as part of Meta’s efforts to combat false news and misinformation on its News Feed. (Read more about our partnership with Meta, which owns Facebook and Instagram.)

The reel’s caption said “the EMA,” which appears to refer to the Emergency Management Agency, encouraged people to stock up on supplies before the eclipse and warned that cell phone signals might be lost due to activity overload. The post’s language is similar to more ominous claims we have checked warning people of alleged emergencies and doomsday scenarios. 

The post’s caption also linked to a March 13 IFLScience article, which included important context that the Facebook post lacked. 

In the article’s first sentence, it said that Emergency Management Agency officials in Lorain County, Ohio, in the path of totality, encouraged residents “to stock up on food, water, and fuel ahead of the total eclipse on April 8, with an influx of visitors to the area expected.”

This image from the NASA Eclipse Explorer website shows the path of the April 8, 2024 total solar eclipse over North America. (NASA via AP)

The information was provided in anticipation of large crowds, not because of complications from the eclipse.

Lorain County officials posted “resident preparation tips” focused on ways to avoid problems caused by a visitor spike. 

Because the eclipse will come on a Monday, “it is likely that many visitors will come in over the weekend and stay in the area,” Lorain County officials wrote. 

Visitors could cause “difficulty accessing food, gas, and supplies due to traffic issues” and “possible loss of cell phone signal due to system overload,” officials wrote. They advised residents to buy gasoline and stock up on food and drinks before April 6 and also said residents should avoid traveling that weekend. 

PolitiFact found other examples of state and local officials in cities along the eclipse’s path of totality warning residents to prepare for crowds of visitors. Similar warnings were issued in 2017, before the Aug. 21 total solar eclipse that year.

This multiple exposure photograph shows the progression of a partial solar eclipse over the Gateway Arch in St. Louis on Aug. 21, 2017. (AP)

In 2017, the Federal Emergency Management Agency warned of “gridlocked traffic conditions,” “travelers stopped on roadways,” “distracted driving during the event,” and “limited cell phone service” or “limited gasoline availability” because of the large gatherings along the path of totality. News organizations at the time reported possible phone service outages because crowds might overtax networks, not because the eclipse would affect phone service. 

In 2017 and now, suggestions to buy food and fuel early focused on complications from large crowds — not from the rare solar event itself. For people living outside the path of totality, we did not find similar eclipse guidance then or now. 

Our ruling

A Facebook reel’s text said emergency officials warn to “stock up on food, water, and fuel ahead of the eclipse.” 

The reel’s lack of specificity could have had people thinking that national emergency management officials were warning the April eclipse might cause complications on its own. We found that some local and state officials are encouraging people living along the eclipse’s path of totality to brace for complications caused by crowds of out-of-town eclipse viewers. 

The post’s claim is partially accurate, but omits important details. We rate it Half True.

PolitiFact Researcher Caryn Baird contributed to this report.

Recent reporting claimed the House select committee that investigated the Jan. 6, 2021, attack on the U.S. Capitol hid evidence related to the White House’s response that day.

Conservative news outlets, including The Federalist, The Blaze and Newsmax, reported that the select committee “suppressed testimony” from former Deputy White House Chief of Staff Anthony Ornato that proves former President Donald Trump pushed for 10,000 National Guard troops at the Capitol.

Facebook posts also shared news articles making this claim. These posts were flagged as part of Meta’s efforts to combat false news and misinformation on its News Feed. (Read more about our partnership with Meta, which owns Facebook and Instagram.)

(Screengrab from Facebook)

We reached out to these news outlets for comment but did not hear back before publication.

The news outlets cited a transcript of a Jan. 28, 2022, interview with Ornato that Rep. Barry Loudermilk, R-Ga., who leads the House Administration’s Subcommittee on Oversight, publicly released March 8. House Republicans have reopened an investigation into the Jan. 6, 2021, attack because they deemed the select committee’s work “incomplete.”

“The Select Committee’s failure to disclose this transcript is additional evidence that the Select Committee only released evidence that fit their narrative,” the Oversight Subcommittee said in its first report, released March 11.

The Jan. 6 select committee, composed of seven Democratic members and two Republican members, concluded its work in December 2022 with a 845-page final report. During its 18-month investigation, the select committee held 10 public hearings, interviewed more than 1,000 witnesses and collected more than 1 million documents.

Ornato’s January 2022 interview transcript was not publicly released until recently for security reasons, Reps. Jamie Raskin, D-Md., and Zoe Lofgren, D-Calif., told PolitiFact. Both Raskin and Lofgren served on the select committee.

The transcript aligns with the select committee’s conclusion that Trump didn’t order the deployment of 10,000 National Guard troops before or during the Jan. 6, 2021, attack. It provides no evidence to negate this finding.

We previously fact-checked claims that Trump requested thousands of troops and Democrats rejected this request. These claims are False. There’s no evidence Trump gave this order.

Why Ornato’s transcript wasn’t publicly released until now

As part of the select committee’s investigation, the committee members and staff interviewed Ornato and Secret Service personnel. Ornato left the Secret Service in 2019 to become Trump’s deputy chief of staff and was still in that role Jan. 6, 2021.

The select committee was “obligated to return certain Secret Service transcripts,” including this Ornato transcript, to the Department of Homeland Security “for redaction of sensitive security information before public release,” Raskin told PolitiFact.

Lofgren also said this was the case.

This obligation was documented in a publicly available letter the select committee sent to the Department of Homeland Security in December 2022.

The letter said the transcribed interviews with Secret Service personnel were “not intended for public release” at the Secret Service’s direction. Therefore, the select committee said it summarized facts from these interviews for its final report without revealing the Secret Services’ operational details or its agents’ personal information.

One of Ornato’s transcribed interviews, conducted in November 2022, was among the select committee’s publicly released materials. The committee said in the letter that this transcript was published because it “addressed a range of intelligence information important to the Committee’s conclusions about January 6th.”

The committee noted in the letter that Ornato was not a functioning Secret Service agent at the time and it significantly redacted the November 2022 transcript to address the Secret Service’s security concerns.

The Oversight Subcommittee said in its report that Loudermilk sent letters to the Department of Homeland Security in August 2023 and January 2024 requesting all of the select committee’s interview transcripts. The department said it possessed 12 transcripts and gave the subcommittee six of them. The others have not been released because the department’s review is not complete, the Oversight Subcommittee’s report said.

What the Ornato transcript says about Trump’s National Guard comments

In the 153-page transcript of Ornato’s January 2022 interview, mentions the National Guard multiple times. Ornato said he recalled Trump floating the number 10,000, but said the president never gave the order.

First, Ornato was asked whether he knew if Trump had asked Jan. 4, 2021, for 10,000 National Guard troops to be deployed Jan. 6, 2021.

Ornato responded, “I was not aware of that.”

Soon after, Ornato was asked again whether he recalled any discussions before Jan. 6, 2021, about “having 10,000 troops or any other number of troops” to deploy.

Ornato said he recalled overhearing a phone conversation a day or two before the Capitol attack between Trump’s Chief of Staff Mark Meadows and Washington, D.C. Mayor Muriel Bowser in which Meadows asked if Bowser needed more guards. Ornato said he had heard only Meadows’ side of the conversation.

“I remember the number 10,000 coming up, you know, the president wants to make sure that you have enough. You know, he is willing to ask for 10,000,” Ornato said. “I remember that number. Now that you said it, it reminded me of it.”

Ornato said he did not hear any discussion of 10,000 troops after this. Ornato said Meadows asked the Defense Department to set up a quick reaction force at Joint Base Andrews, just outside of Washington, D.C., in case additional National Guard troops were needed.

When asked, Ornato said he knew of no order to deploy National Guard troops on the morning of Jan. 6, 2021.

“And apart from that conversation that you overheard with Mayor Bowser and Chief Meadows, you did not learn of any other additional efforts or (an) order regarding 10,000 troops?” Senior Investigative Counsel Soumyalatha Dayananda asked Ornato.

Ornato responded, “No, not 10,000 troops, no.”

Our ruling

Conservative news outlets and social media posts claimed the Jan. 6 select committee “suppressed testimony” from Ornato that proves Trump pushed for 10,000 National Guard troops at the Capitol.

The select committee did not release this transcribed interview with Ornato because it was under the Department of Homeland Security’s review. As a former Secret Service member, Ornato’s transcript needed to be reviewed for potential security concerns.

The select committee considered Ornato’s and other Secret Service testimony when publishing its final report on the Jan. 6, 2021, attack. The report concluded that Trump did not directly order deploying 10,000 National Guard troops before or during the attack. Ornato’s interview transcript does not negate this finding.

We rate the claim that the Jan. 6 select committee “suppressed testimony” involving 10,000 National Guard troops False.

Political turmoil in Haiti has reached crisis levels and the nation’s prime minister agreed to resign under pressure from gangs. Misleading social media posts are adding to the confusion.

An Instagram post shared what looked like a screenshot from CNN’s official X account: “Kenyan present William Ruto reportedly called president Joe Biden to organize the purchase of ‘several dozen’ WW2 era amphibious landing crafts for a naval invasion of the Haitian coast,” read the March 14 post, which appeared to misspell Ruto’s title. 

According to the image, the statement appeared on CNN’s X account at 10:25 a.m. March 12. 

This post was flagged as part of Meta’s efforts to combat false news and misinformation on its News Feed. (Read more about our partnership with Meta, which owns Facebook and Instagram.)

The post is fabricated, CNN spokesperson Emily Kuhn, told PolitiFact. CNN’s official X account shows no evidence of the post and it’s not in CNN’s X account archives. The screengrab shows CNN’s avatar as a red circle while the real account has a square logo as its avatar.

Kuhn said the news the faux post contained is “not something CNN reported or posted on X.” 

CNN X posts typically contain links to CNN articles, but the Instagram screenshot includes no links. We searched Google and Nexis news archives and found no CNN articles about Ruto and Biden planning to invade Haiti. 

We also searched other publications for news of Kenya seeking to use U.S. vessels to invade Haiti but found no credible reports. 

Kenyan President William Ruto said March 13 that his country would lead a United Nations-backed multinational police force to help reduce gang violence in Haiti after a transitional presidential council installs an interim president. The police force would consist of at least 1,000 police officers, but the report didn’t mention naval vessels. 

In 1915, the U.S. Marine Corps invaded Haiti to address political instability. The U.S. occupied the Caribbean nation until 1934.  

We rate the claim that CNN posted on X that Ruto called Biden to organize a “naval invasion of the Haitian coast” False.

Una publicación en redes sociales hizo lo que parecía ser un anuncio sorprendente sobre la carrera presidencial de 2024.

“Últimas noticias: Donald Trump ha declarado que se retira de la carrera presidencial”, dice el narrador en un video en inglés en TikTok, el cual fue visto más de 161,000 veces antes de que TikTok lo removiera de su plataforma. 

Por supuesto que Trump no se ha retirado de la carrera de 2024. El 12 de marzo, el aseguró suficientes delegados para obtener la nominación presidencial republicana. 

Además de la afirmación falsa, el video tenía otro elemento distintivo: El audio parecía ser hecho con inteligencia artificial.

PolitiFact identificó varias cuentas en TikTok esparciendo narrativas falsas sobre las elecciones de 2024 y Trump a través de nuestra colaboración con TikTok para combatir contenido inauténtico, desinformativo o falso. (Lea más sobre nuestra colaboración con TikTok).

Videos en YouTube también hicieron afirmaciones similares. Estos videos, los cuales también parecían tener audio generado por IA, fueron compartidos en Facebook y X con menos participación.

La IA Generativa es un término amplio que describe cuando computadoras crean nuevo contenido, como texto, fotos, videos y audio, identificando patrones en datos existentes. Este año, los Estados Unidos y más de otros 50 países tienen elecciones nacionales, y es probable que la IA generativa juegue un papel destacado. En febrero en Indonesia, caricaturas generados por IA ayudaron a rehabilitar la imagen de un ex general militar que estaba relacionado a abusos de derechos humanos, quien ganó la elección presidencial en el país. 

Avances recientes en la AI generativa han hecho más difícil determinar si el contenido en internet es real o falso. Hablamos con expertos sobre cómo esta tecnología está cambiando el panorama de la información y como saber si un audio fue generado por IA. A diferencia de los vídeos, no hay anormalidades visuales que puedan dar señales de manipulación.

Cuando contactamos a TikTok para una declaración, un vocero dijo, “Para aplicar nuestras políticas de desinformación dañina, detectamos el contenido engañoso y lo enviamos a nuestros colaboradores verificadores para una evaluación factual”. Una vez alertados sobre el contenido señalado en esta historia, TikTok lo removió de su plataforma. 

Análisis del audio generado por inteligencia artificial

Le pedimos a expertos en inteligencia artificial generativa que analizaran cinco videos en TikTok que hicieron afirmaciones falsas y engañosas en múltiples cuentas para determinar si habíamos acertado en que el audio había sido generado con IA. 

Hafiz Malik, un profesor de ingeniería eléctrica y computación que estudia “deepfakes” en la Universidad de Michigan-Dearborn, dijo que sus herramientas de detección de IA clasificaron cuatro de los videos como “sintético”, o audio generado por IA. El quinto fue señalado como de “baja confianza”, lo cual Malik dice que significa que algunas partes fueron etiquetadas como “deepfake”, pero otras no. Los videos “deepfake” son sofisticados, generados por computadora y dan la impresión que alguien está diciendo o haciendo algo, aun cuando esa persona no dijo o hizo eso.

Siwei Lyu, un profesor de informática e ingeniería de la Universidad de Buffalo que se especializa en análisis forense de medios digitales, dijo que sus algoritmos de detección de inteligencia artificial también determinaron que la mayoría de los cinco videos usaron audio generado por IA.

Videos en TikTok hacen afirmaciones falsas y escandalosas

Trump fue el foco de estas cuentas en TikTok; varias de las cuentas utilizaron fotos del expresidente como sus fotos de perfil.

(Capturas de pantalla de TikTok)

Estas cuentas siguieron un esquema similar: titulares llamativos a menudo mostrados sobre fondos rojos; videos que contienen clips o fotos de figuras famosas, incluyendo a Trump y el juez de la Corte Suprema Clarence Thomas; y a un narrador oculto.

Algunos de los videos hicieron declaraciones falsas sobre el bienestar de personas de alto perfil — que Trump había tenido un ataque al corazón y que la secretaria de Justicia de Nueva York Letitia James fue hospitalizada por heridas de balas. James demandó a Trump en 2022 acusándolo de inflar fraudulentamente su valor neto; un juez dictó en febrero que Trump debe pagar una multa de $454 millones. 

Otro video mostraba un texto que decía, “Juez de la Corte Suprema Clarence Thomas se une a otros jueces para remover al candidato de la carrera de 2024”.  Esto parece referirse engañosamente al caso de la Corte Suprema sobre la elegibilidad de Trump para estar en la boleta de votación, en el cual Thomas y otros jueces dictaron por unanimidad que los estados individuales no pueden suspender a candidatos presidenciales, incluyendo a Trump, de la boleta. 

(Capturas de pantalla de TikTok)

Todas estas cuentas en TikTok fueron creadas en los últimos meses. La que parece más antigua tiene videos de noviembre de 2023; la más nueva comenzó a publicar contenido el 1 de marzo. Tres de las cuentas tenían nombres de usuarios designados por TikTok como “usuario” seguido por 13 números.

Estas cuentas publicaron colectivamente cientos de videos y obtuvieron miles de vistas y reacciones de me gusta y seguidores, antes de que TikTok removiera las cuentas y videos. 

También encontramos videos en YouTube haciendo las mismas declaraciones falsas como los videos de TikTok. Estos videos imitaron el formato de los videos en TikTok: titulares sensacionalistas, fotos de Trump, y audio que suena generado por IA.

(Capturas de pantalla de YouTube)

La mayoría de los videos en YouTube fueron vistos cientos o miles de veces. Las dos cuentas de YouTube que publicaron los videos fueron creadas en 2021 y 2022. Estas cuentas acumularon decenas de miles de seguidores antes de que contactaramos a YouTube por una declaración y la compañía removiera las cuentas de la plataforma. 

“Contenido involucrando prácticas engañosas que confunden y se aprovechan de la comunidad de YouTube no es permitido. Como tal, cancelamos los canales de On Point Juicy News and THE BREAK TIME por violar repetidamente nuestras politicas de spam, practicas engañosas y estafas”, dijo Ivy Choi, una vocera de YouTube. 

Antes de que YouTube eliminara los videos, estos fueron compartidos en otras plataformas de redes sociales, incluyendo Facebook y X, donde pocas personas le dieron me gusta o las vieron. (Lea más sobre nuestra colaboración con Meta, propietaria de Facebook e Instagram).

¿Cómo contribuye la IA generativa a la desinformación en internet?

Los expertos en desinformación afirman que las nuevas generaciones de IA generativa están facilitando la creación y difusión de contenidos engañosos en redes sociales. Además, crear audio generado por IA suele ser más barato que el video.

“Ahora, cualquiera con acceso a internet puede disponer del poder de miles de escritores, técnicos de audio y productores de video, de forma gratuita y con sólo pulsar un botón”, afirma Jack Brewster, un editor en NewsGuard, una empresa que rastrea la desinformación en internet.

“En las manos adecuadas, ese poder puede utilizarse para el bien”, dijo Brewster. “En las manos equivocadas, ese poder puede utilizarse para contaminar nuestro ecosistema de información, desestabilizar las democracias y socavar la confianza pública en las instituciones”.

​​NewsGuard informó en septiembre de 2023 de que se estaba utilizando tecnología de voz de IA para difundir teorías conspirativas en masa a través de TikTok. NewsGuard reportó que “identificó una red de 17 cuentas de TikTok que utilizaban software de texto a voz de IA para generar videos que avanzaban afirmaciones falsas y sin fundamento, con cientos de millones de visualizaciones.”

Un estudio de la Universidad de British Columbia de 2023 que utilizó datos de TikTok descubrió que la tecnología de conversión de texto a voz mediante IA simplificaba la creación de contenidos, motivando a los creadores de contenidos a producir más videos.

Los autores del estudio, Xiaoke Zhang y Mi Zhou, dijeron a PolitiFact que el aumento de la productividad significa que la IA generativa “puede explotarse deliberadamente para generar desinformación a bajo costo”. 

La tecnología también puede ayudar a los usuarios a ocultar su identidad, lo que puede “disminuir su sentido de la responsabilidad a la hora de garantizar la exactitud de la información”, señalaron Zhang y Zhou.

TikTok exige a los usuarios que etiqueten los contenidos que contienen imágenes, videos o audio generados por IA “para ayudar a los espectadores a contextualizar el video y evitar la posible difusión de contenidos engañosos.” Las directrices de la comunidad de TikTok prohíben el “contenido inexacto, engañoso o falso que pueda causar un daño significativo a las personas o a la sociedad”.

Ningún video de TikTok que revisamos tenía esta etiqueta de IA generativa, aunque algunos incluían etiquetas para aprender más sobre las elecciones de EE.UU. Brewster dijo que NewsGuard también observó que muchos usuarios de TikTok se saltaban esta política sobre la identificación de contenidos generados por IA.

Las directrices de la comunidad de YouTube no permiten “contenido engañoso que genere riesgos graves de daño flagrante”. YouTube exige la divulgación de la publicidad electoral que contenga “materiales alterados o generados digitalmente”. La compañía dijo en diciembre que planea ampliar esta divulgación de IA generativa a otros contenidos.

Cómo detectar audio generado por IA

Los expertos dicen que las herramientas de detección de IA existentes son imperfectas. Añaden que, a medida que mejoran las herramientas de detección, también lo hace la tecnología de IA generativa.

El audio generado por IA carece de señales visuales que se observan en videos,  como los movimientos de la boca que no están sincronizados con el audio o los rasgos físicos distorsionados.

Sin embargo, hay formas de identificar el audio generado por IA.

Malik, profesor de la Universidad de Michigan, dice que hay que prestar atención a las anomalías en el tono vocal, la articulación o el ritmo.

“(Las voces generadas por IA) carecen de emociones. Carecen de las subidas y bajadas de volumen que se producen normalmente al hablar”, dijo Malik. “Son bastante monótonas”.

Brewster también aconsejó que las “viejas tácticas” siguen siendo la mejor manera de evitar la desinformación generada por la IA. Por ejemplo:

• Contrastar la información con la de otros sitios

• Estar atento a los errores gramaticales y a las expresiones extrañas

• Buscar los nombres de los autores para ver si han compartido información falsa en el pasado.

Una versión de este artículo fue escrito originalmente en inglés y traducido por Maria Briceño y Marta Campabadal.

Read a version of this article in English.

Lea más reportes de PolitiFact en Español aquí.

---

Debido a limitaciones técnicas, partes de nuestra página web aparecen en inglés. Estamos trabajando en mejorar la presentación.

Videos of people fleeing the sounds of gunshots, images of buildings on fire and headlines of political unrest in Haiti have increased over several weeks.

Violence escalated after a coalition of gangs began attacking Haiti’s capital, Port-Au-Prince, forcing acting Prime Minister Ariel Henry to resign. He’d assumed the role after President Jovenel Moïse was assassinated in 2021.

Amid the real images of chaos lurks a gruesome, unfounded narrative about a culture of cannibalism in Haiti. 

The Daily Star, a British tabloid, helped kick-start the panic with a March 5 story that said Haitian “cannibal gangs” are “eating people they’ve killed,” citing an anonymous Haitian journalist.

Conservatives with verified accounts and large platforms, including Elon Musk, the owner of X, amplified the unverified claims on social media and warned of an “invasion” of Haitian cannibals in the U.S.

“There are cannibal gangs in Haiti who abduct and eat people,” Ian Miles Cheong, a right-wing commentator, posted on X on March 6. “Reminder that these people are now illegally entering the U.S. en masse,” he added.

Online posts in Spanish used the same videos with similar comments.

The online rumors coincide with  former President Donald Trump comparing migrants to Hannibal Lecter, the serial killer and cannibal character in “The Silence of the Lambs,” during presidential campaign rallies.

These posts, however, do not offer evidence that Haitians in the country, or Haitians seeking asylum in the U.S., are practicing cannibalism. The videos Musk and others have shared include outdated videos that either couldn’t be verified or were taken in other countries. 

The State Department told PolitiFact that it has received no credible reports about cannibalism in Haiti, and experts who study the country said they’ve seen no sign that cannibalism is prevalent among gangs or the population at large. 

“There is ZERO evidence that any cannibalism is taking place now, or ever has occurred in the past, in Haiti,” Marlene Daut, a Yale University professor and Caribbean historian, wrote in an email. “There is similarly absolutely no evidence that Haitians migrants to the U.S. seeking asylum are ‘cannibalists.’ Those claims are, of course, patently racist and stem from longstanding efforts by white supremacists to dehumanize Haiti and Haitians.”

What the videos showed

Several videos shared to support the cannibal claims are not connected to Haiti or the current unrest.

For example, one video’s depiction of roasting corpses was footage of Halloween decorations at a Chinese theme park in 2018, Snopes reported. 

Another account shared footage from a 2017 CNN documentary series that interviewed a member of a Hindu sect that includes postmortem cannibalism in its religious practices. The years-old interview is from India. 

We could not verify another video that appears to show cannibalism. NBC News reported the video had been online “for two years or longer.” Some X users claimed the clip showed a battle between two gangs in Haiti’s Artibonite Valley. An extensive Google and Nexis search of news articles yielded no evidence of cannibalism in that area. A United Nations report of violence in the region detailed accounts of “extreme brutality” by gangs, but did not mention cannibalism.

International groups and governments have documented violence in Haiti for decades, publishing extensive reports of the gangs’ violent tactics, which include extortion, indiscriminate murders, kidnappings and rape. PolitiFact found no descriptions of cannibalism.

One of Haiti’s most prominent gang leaders, Jimmy Chérizier, goes by the alias “Barbecue” — which is boosting the cannibalism talk. Chérizier told The Associated Press in 2019 that he got his nickname as a child because his mother was a street vendor who sold fried chicken. Although there is plenty of information about Chérizier’s violent tactics — the United Nations sanctioned him over human rights violations — no sources include cannibalism.

History of cannibalism claims in Haiti

Narratives painting Haiti as a barbarous country of cannibals are not new.

“These kinds of claims have a long and deep history, and they tend to say more about the people who make them than they do about Haiti itself,” said Laurent Dubois, a University of Virginia historian. “There is probably no country in the world that has had more misrepresentations projected onto it than Haiti.”

The origin of “cannibal” traces back to Italian explorer Christopher Columbus’ voyages in the late 15th century, Dubois said. It came up “out of a misheard indigenous term (Columbus)  encountered in 1492 when he disembarked in the north of what is now Haiti,” Dubois said.

Stories of cannibalism in Haiti have circulated since.

In 2013, historian Mike Dash detailed the story of an 1860s trial over a child’s murder and cannibalism for Smithsonian magazine. The case “helped define attitudes toward the nation and the (Vodou) ​​religion ever since,” Dash wrote. Haitian Vodou, rooted in a blend of Catholicism and Western and Central African spirituality, was influenced by escaped slaves in the 17th century who wanted to unite under a common spiritual identity. 

Neither Haitian culture nor island religious practices support “anything similar to cannibalism,” said Athena Kolbe, a social work professor at Barry University and the Institute of Social Work and Social Science in Haiti.

Kate Ramsey, a University of Miami history professor, said, “Such tropes go back to the nineteenth century when, following the Haitian Revolution, colonialist and anti-abolitionist detractors sought to discredit Black self-governance in Haiti through claims about the decline of ‘civilization’ in the second independent nation in the hemisphere and the first to permanently abolish slavery.”

Experts pointed to the 19-year U.S. occupation of Haiti between 1915 to 1934 as another flashpoint. U.S. forces were accused of committing atrocities against Haitian people, including shooting civilians, forcing labor and enforcing martial law for long stretches of time.

“The proliferation of negative images about Haiti’s culture played a role in justifying the abuses carried out by the U.S.,” the University of Virginia’s Dubois said. 

Unfounded claims of cannibalism in Haiti foment fear of Haitians, especially of ones fleeing brutal violence.

There is a “deep set of unconscious ideas about Haiti that have been developed over a long history,” Dubois said. And people should be aware of it “to avoid just falling into recycling old and misleading stereotypes.”