12 Dec 2025 | Americas, Brazil, News
Brazilian left-wing influencer Thiago Torres, best known as Chavoso da USP (roughly translated as the University of São Paulo’s swaggy chav), has faced increasing political persecution in the last months. This reached international levels last month when Thiago’s main Instagram profile, with more than one million followers, was taken down by Meta.
Thiago then started using an old backup Instagram account with 385,000 followers, which was also taken down after allegations that it had been created to circumvent the previous block. Arbitrarily and without possibility of appeal, Meta blocked all access to his accounts and is set to permanently delete their content. A warning on Instagram said that the account “does not follow Community Standards” although the company did not specify which specific rules had been breached. Even after a preliminary injunction was issued on the morning of 20 November that forced Meta to return Thiago’s main account under threat of a fine, three other accounts were taken down later that same evening.
By maintaining the block on the influencer, Meta is involved in yet another case of big tech insubordination to Brazilian justice according to politicians. Federal Congresswoman Sâmia Bomfim, from PSOL (Freedom and Socialism Party), classified the event as a “direct attack on freedom of speech and the work of those who denounce injustices within Brazil.” Thiago sees it as “an offensive against progressive, mainly radical, left-wing voices”.
This is not the first time Meta has taken down accounts with large numbers of followers linked to the Brazilian left. In August this year, historian and influencer Jones Manoel, former candidate for governor of Pernambuco with the PCB (Brazilian Communist Party) and the Brazilian influencer with the most growth on the platform since June, was arbitrarily banned from Instagram. In October, activist and comedian Tiago Santineli also had his 850,000 followers account blocked, following online comments about the death of Charlie Kirk.
Since 9 December 2025, members of parliament from PSOL, PT (Worker’s Party), and left-wing news outlets have reported that their profiles “don’t appear in searches, can’t be tagged, and [that] their reach has plummeted in an orchestrated manner”, according to Federal Congresswoman Fernanda Melchionna. This is known as shadow-banning.
The bans follow a dispute between big tech companies and the left wing government of Luiz Inácio Lula da Silva which dates from January 2025, when Brazil’s Attorney General’s Office sent an extrajudicial notification to Meta because of the company’s decision to stop using independent fact-checkers. The concern was that this would further exacerbate the problem of “fake news”, which became prevalent in the 2018 and 2022 election processes, particularly on the part of the Brazilian right wing. A major dispute between the Brazilian judiciary and Elon Musk’s X also took place last year, resulting in the social network being blocked in the country until Musk complied with court orders.
The regulation of big tech companies – largely similar to what the EU has instigated – is considered by the current government as a matter of national sovereignty. In July, President Trump sent a letter to Brazil’s president, Lula, imposing a massive 50% tariff that rendered the export of a range of Brazilian products to the USA unfeasible. According to the letter, the measure came as retaliation for the sanctions against big tech and in support of former president Jair Bolsonaro, a representative of the Brazilian far right and ally of Trump who was convicted for attempted coup d’état.
In his speech at the UN General Assembly in September, President Lula said that “even under unprecedented attack, Brazil chose to resist and defend its democracy. There is no justification for the unilateral and arbitrary measures against our institutions and our economy. The aggression against the independence of the judiciary is unacceptable.”
It is not only in Brazil that US intervention in favour of big tech been felt. Back in January, Meta’s CEO Mark Zuckerberg clearly stated on the Joe Rogan Experience Podcast that “the US government has a role in basically defending [big tech] abroad”.
In the same week that Brazil hosted COP30 and witnessed the preventive arrest of Bolsonaro, the suspension of five accounts belonging to a left-wing influencer shows that big tech might also have a role in defending the US government’s interests abroad in Brazil.
Researchers like the Brazilian academic Walter Lippold denounce what they call “digital colonialism”, the interconnection between imperialist interests and big tech. To Brazilian sociologist Sérgio Amadeu, “online social networks and platforms controlled by big tech companies are geopolitical structures increasingly aligned with the far right.” In June, at seminars held by Bolsonaro’s right-wing Liberal party (PL), executives from Meta gave workshops teaching how to use AI and achieve greater reach on the platform.
Born and raised in Brasilândia, an outlying neighbourhood of São Paulo, Thiago Torres first rose to prominence as a social sciences student at the University of São Paulo.
Ranked many times as the best university in Latin America, the University of São Paulo subscribes to a national public education project aimed at social development. Despite this, USP remains elitist in the social and racial makeup of both its faculty and students. Thiago spoke about the way this composition shaped the production of knowledge within USP, and used his platform to share social theory with a wider public.
Now graduated and a teacher, Thiago has become known for denouncing cases of political corruption and police violence. Overtly anti-capitalist and anti-imperialist, it’s not surprising that his head is wanted by public officials and companies who benefit from the country’s social division.
In August this year, Thiago was called to testify in the controversial CPI dos Pancadões, a parliamentary commission inquiring into street funk parties. Under the pretext that they disturb public order, it is common for the military police to raid pancadões, using extreme violence and murdering the young people present, many of whom are from racial minorities and come from lower social strata.
Thiago’s account dedicated to police violence, @fim.da.pm (“End the Military Police”), is among those blocked by Meta. The company had until 28 November to return the influencer’s main account, but this didn’t happen.
“Instagram will face a daily fine for each day it fails to comply [with the judiciary decision]”, Thiago explained. “But it’s a relatively small fine for them, so it’s possible they might disregard the court order.” Unfortunately, this seems to be the case.
4 Dec 2025 | Iran, Middle East and North Africa, News
The line between fact and fiction often overlaps in Jafar Panahi’s films.
Take Taxi Tehran from 2015 for instance. The film takes place inside a cab with three hidden cameras. Panahi, an internationally acclaimed award-winning Iranian director, plays himself. He just so happens to be driving a taxi around the Iranian capital. What initially seems like an improv documentary eventually turns out to be a satirical conceit. Namely: the director is using the safe space of a private car to freely discuss what would ordinarily be off limits to discuss publicly in the Islamic Republic of Iran.
Among the passengers that Panahi picks up is the Iranian human rights lawyer, Nasrin Sotoudeh. Over the last 15 years she has been imprisoned twice in her native country. Her last stint was for defending women prosecuted for appearing in public without a hijab. Sotoudeh’s husband, Reza Khandan, is also now serving a three-and-a-half year prison sentence for voicing public opposition to Iran’s compulsory hijab laws. In Taxi Tehran, Sotoudeh speaks about defending human rights and free speech in a theocratic-totalitarian-police state. “First they mount a political case,” Sotoudeh explains. “They beef it up with a morality charge, then they make your life hell.”
In that same scene, Sotoudeh notices the director looking at his back window.
“Looking for someone?” she asks.
“I heard a voice … I thought I recognised my interrogator,” Panahi replies.
Sotoudeh mentions how her clients often say this. “They want to identify people by their voices,” she says. “Advantage of blindfolds.”
“This reference in Taxi [Tehran] to prisoners hearing sounds is a communal experience shared by all prisoners of conscience,” Panahi told Index from Los Angeles, via a Farsi translator. “In my current film I wanted to talk about a [similar] experience. This time, however, the sound is coming from a disabled [prosthetic] leg, which becomes the moving engine of the film.”
Panahi’s latest movie, It was Just an Accident, won the Palme d’Or at this year’s Cannes Film Festival and opens in UK cinemas this coming Friday, 5 December. The story begins in a mechanics’ workshop, where a man named Vahid is convinced he has just encountered Eghbal – a prison inspector who once caused him great pain and suffering. Vahid hears Eghbal before he sees him. He can never forget the eerie squeaking sound Eghbal’s prosthetic leg makes in motion. He remembers it from prison, where “Peg Leg” was known as a sadistic torturer. The traumatised Iranian mechanic later kidnaps Eghbal and even considers killing him. But has he got the right man? To tease out his doubts, Vahid rounds up a group of former prisoners to seek their advice.
What follows is a brilliant farcical black comedy-road trip movie. Despite the light-hearted banter, the film poses two serious ethical questions. One, how far will an individual – or a group – go to seek revenge on former enemy? Two, at what point does revenge violence make the victim the victimiser?
It was Just an Accident has been selected by France as its official nomination for the Academy Awards this coming March. It may be Panahi’s most overtly political film to date. But the 65-year-old Iranian moviemaker disagrees.
“I don’t make political films, which typically tend to divide people into good and bad,” he insists. “I make social films, where everyone is a human being.”
The film’s script was inspired from several conversations Panahi had with inmates he befriended while serving time in Tehran’s notorious Evin Prison. In March 2010, Panahi was convicted by a Revolutionary Court in Iran of propaganda for his film-making and political activism. He subsequently spent 86 days behind bars, he explained: “For the first 15 to 20 days I was in a small cell in solitary confinement, where I was interrogated.”
That same year, the Iranian regime handed Panahi a 20-year ban that forbade him from directing films or writing screenplays. “This [censorship] I experience presents many challenges to keep making films, but a social filmmaker is inspired by the circumstances in which they live,” said Panahi. “If I lived in a freer society what would inspire me? I don’t know.”
Despite the ban, Panahi is a prolific filmmaker who never stops creating. Many of his films have focused on the complications of making films with a state-imposed censorship hanging over his head. They include the ironically titled, This Is Not a Film (2011), which was smuggled out of Iran on a USB stick concealed inside a birthday cake, and No Bears (2022).
No Bears has two stories in it. The first is about migrants heading off to Europe and the second is about Panahi, who is stuck back in Iran, as his film crew attempt to complete the film they are shooting just across the border in Turkey. The film won the Special Jury Prize at the 2022 Venice Film Festival. Panahi was unable to collect the prize. He was then back in Evin prison, after a Tehran court ruled he must serve the six-year sentence he was handed more than a decade before for supporting anti-government demonstrations.
“According to the law [in Iran] if a sentence is issued but not gone into effect for ten years, it should not be executed,” said Panahi. “However, [the regime] said that this is not true about political prisoners. They were lying though.”
During this second prison term Panahi was in a public ward with 300 or so prisoners, of whom roughly 40 were prisoners of conscience, he said: “On that occasion I did not face interrogations or solitary confinement, which meant I could speak and listen to the prisoners’ stories.”
Panahi remained in prison until the following February. After his release, he noticed many changes in Iranian society. The previous September, a 22-year-old Kurdish woman, Mahsa Jîna Amini, was beaten to death in Tehran by Iran’s so-called morality police after being accused of defying the country’s hijab rule. The state sanctioned homicide inspired the Woman Life Freedom uprising, which saw an estimated two million take to the streets across Iran. Many ripped and burnt posters of their political leaders, while others openly chanted, “Death to the Islamic Republic!” Iranian security forces, meanwhile, responded by killing hundreds of protesters.
“The history of the Islamic Republic [will eventually] be divided into before and after the timeline of this movement,” said Panahi. “The impact has been enormous and even made its way into cinema.”
Specifically, Panahi was referring to the fact that many women who appear in It Was Just an Accident – including actors and extras – are not wearing the hijab. “Much of what you see in the background of the film is people being filmed as they are in daily life in Iran today,” said Panahi. “For example, one woman who agreed to be in the film as an extra said to me: ‘If you are going to force me to wear the hijab, I am not going to do that.’ I told her: ‘You appear as you wish’.”
It’s not a view the authorities in the Islamic Republic endorse. Just days after Index spoke to the Iranian director, he was sentenced in absentia to one year in prison and a travel ban over “propaganda activities” against Iran. The news was broken via the French news agency, Agence France Presse (AFP), who cited Panahi’s lawyer, Mostafa Nili, as a source.
At the time of writing, Panahi remains outside Iran. Prior to news of his new prison sentence being issued, Panahi told Index he could not imagine living somewhere in which he has only a touristy outlook and superficial understanding of the people and culture: “I have lived in Iran for 65 years and I make films about Iranians. I don’t want to stop making films because life without cinema has no meaning to me.”
“[In Iran] when you work you will have problems as a filmmaker there and anywhere in the world the Iranian authorities can get their hands on you,” Panahi concluded. “But you accept this is the price to be paid, and you get through what you have to in order to make the film you want to make.”
13 Nov 2025 | Americas, Asia and Pacific, Europe and Central Asia, Germany, India, News, Singapore, United States, Volume 54.03 Autumn 2025
This article first appeared in Volume 54, Issue 3 of our print edition of Index on Censorship, titled Truth, trust and tricksters: Free expression in the age of AI, published on 30 September 2025. Read more about the issue here.
We are right to be concerned about what artificial intelligence is doing to our present and what it might do to our future. But I am more worried about what it can do to our past.
Access to archives is becoming harder. Manuscripts are fragile, and verifying historical evidence contradicting established narratives is time-consuming. Technology makes history seem fun and exciting by enabling direct communication with historical figures and using visuals that simplify the past, but it weakens our understanding. Artificial intelligence reshapes research by digitising archives and analysing data quickly, but it poses risks. AI often lacks academic rigour, drawing from biased sources and oversimplifying complex events. It presents information with unearned authority, spreading errors rapidly. Unlike historians, AI cannot evaluate credibility, weigh accounts or identify gaps in records.
The result is that our history runs the risk of being skewed towards what’s most accessible online: mainstream narratives drawn from popular databases, digitised books, encyclopaedias, widely-read history books and even crowd-sourced portals. Marginalised voices – Indigenous people, minorities, or communities without digitised records – risk being erased further. When AI amplifies the dominant version of events, alternative interpretations fade into obscurity.
For all its potential, AI cannot replace the human historian. Critical judgment, contextual thought and insights, and the ability to navigate conflicting evidence and picking plausible theories remain essential to safeguarding the integrity of the past.
Consider a recent experiment by Indian historian Anirudh Kanisetti. A trained engineer, Kanisetti has written two engrossing books on the Chaulukyas and the Cholas, two medieval-era dynasties of southern India. Despite his background in engineering, or perhaps because of it, he is sceptical of the use of AI in history. He decided to challenge the Goliath. He sought AI assistance and then wrote about his experience on Instagram. He calls AI “language calculators, and not very good ones”. He asked Copilot, Microsoft’s AI tool which uses OpenAI’s ChatGPT, to write an essay based on his work about a medieval regiment in India.
In its references, the tool produced a paper by an academic Kanisetti knew of, but he had never heard of this paper. He called out Copilot, and it immediately admitted it had made up the fact. Its subsequent apology was insincere. Kanisetti then posed another query, instructing Copilot to cite primary sources. This time, the bot confidently misquoted primary sources and, again, admitted the fabrication when he called it out.
Kanisetti is worried because he knows the sources and can find out when AI is lying; other users may not be that knowledgeable and may rely on AI to do the grunt work of checking citations. As he puts it: “Large language models (LLMs) are trained to seem to be helpful, so you think they have value. But they are actually faffing, lying, fabricating … Engagement generates shareholder value.”
Our information ecosystem is being destroyed, he added.
He has grasped the essence of the problem that is bothering many historians in the global south.
The future historian’s skillset
Aanchal Malhotra, who has written a fascinating history about India’s Partition, told me: “This thought of AI rewriting history is new but a terrifying one, because of the scale it can achieve.”
Doyenne of Indian history Romila Thapar told Index: “What AI can do to history is in a sense already being done by what the Hindutva-vadis [right-wing Hindu nationalists] are doing to Indian history. It is an alternative, distorted history that they are propagating. So far, professional historians are dismissing it by showing that there is no reliable evidence to provide proof of what is being stated in the Hindutva version. One can predict that with the availability of AI there can be massive forging of many documents that will be put forward as proof. So training in professional history will require the ability to recognise forged documents, especially if they are documents that are said to belong to earlier times. The historian of today has to be a multidisciplinary person, but the historian of tomorrow will also have to be trained in the technology of verifying documents and publications.”
She wonders if the techniques of historical excavation will need to change, since the authenticity of three-dimensional artefacts gathered from sites will have to be proved. As soon as an artefact is discovered, will it need to be intensively photographed and marked and documented and a tiny sample be analysed? This will make excavation absurdly expensive if the veracity of every object has to be proved, she adds. And she is concerned about how AI’s apparent ability to bring characters from the past back to life would interfere with what we know to be true.
She also worries about what AI can do to historical questioning and education. If students start asking AI to write their papers and tutorials (as many are), they will become masters at writing prompts, but not at drawing their own analysis. They will know how to use a machine, but truth will elude them. “They won’t know why they have written what they have,” she said.
History abounds with examples of misinterpretations, such as the figurine of a horse in Mohenjo-daro in Pakistan, which suggested that the horse had been domesticated as early as the second millennium BC.
Peter Frankopan, who teaches history at Oxford, told Index: “The skill of a historian is to be able to deal with complex materials… The idea of lots of forgeries is as old as writing systems. Information has always been used and manipulated to embed hierarchies, to spread falsehoods and to manipulate people – whether literate or otherwise. So training historians in the present and future simply requires the right skills to be learned, and the discipline of asking the right questions. Clearly, we have some challenges ahead; but there have always been such problems in the past.”
The risks heighten with the seductive charm of AI-generated visual imagery. Manu Pillai, who has written thoughtful and engrossing books on the history of southern India and, more recently, about the making of modern Hindu identity, told Index it was only a question of time before we find sophisticated imagery or videos generated by AI that would make the crudely written WhatsApp messages appear believable. There are twin dangers in India, he said. There is high digital access but weak digital literacy, “which means large segments of people are likely to be misled. We also live in a time when more and more people have an appetite for conspiracy theories and for ‘alt’ facts, and many therefore would be predisposed to swallow some of the material that comes their way”.
AI for good…occasionally
AI does provide notable benefits, such as drawing patterns from vast amounts of data and deciphering handwriting. It helps restore and better understand complex historical texts. The Arolsen Archives use AI to catalogue documents related to Nazi persecution victims. Yad Vashem uses AI to identify unknown Holocaust victims. The USC Shoah Foundation and Illinois Holocaust Museum use AI-driven holographic displays, voice recognition and virtual reality to create transparent, immersive experiences for remembrance and education.
AI’s inability to separate truth from lies is the real danger. As an aggregator, if it finds a particular version of a story cited more often, it is trained to assume it is part of the mainstream discourse. It overemphasises the dominant narrative over alternative views, which reinforces misinterpretations and falsehoods.
Manila-based Singaporean historian Thum Ping Tjin (better known as PJ Thum) is wary. The founder of New Naratif notes that throughout human history, tools have emerged to make our communication and information exchange quicker, and these tools have raised alarms. They have inevitably been co-opted by those with money and power in order to get more money and power. “They have used those tools to present versions of history to further their own goals,” Thum said.
“In many cases, the worst-case scenario you fear already exists,” he told Index. “Southeast Asian governments have long used their resources to fabricate, censor and present their own version of history and their power to enforce it. Singapore has had an entire industrial complex of official historians repeating the ‘official’ version of history. Professional historians who seek to correct the record are treated as public enemies.”
Thum himself underwent hours of gruelling questioning by government officials when he critiqued Singapore’s official interpretation of how it got separated from Malaysia, and the government introduced a new law intended to attack disinformation but which in effect ensured that the dominant, state-approved narrative would prevail.
Alarmed by the attacks on crucial aspects of American history, primarily dealing with race, the American Historical Association Council has issued guidelines based on principles that reinforce the need for historical thinking, reminding us that AI produces texts, images, audio and video, but not truth. It also warned about AI’s tendency to hallucinate and introduce false certainties.
AI’s biggest impact on history is deepfake technology. Erasing inconvenient truths or spreading lies isn’t new – Joseph Stalin famously airbrushed Leon Trotsky out of photos. In the late 1990s, a Malaysian newspaper altered photos to remove Anwar Ibrahim after he fell out with prime minister Mahathir Mohammed.
The problem of AI fakes
Forged documents have long confused historians; for example, Nazi era expert Hugh Trevor-Roper believed the fake Hitler diaries were genuine because of their voluminous content. The fabricated 1903 Protocols of the Elders of Zion fuelled antisemitism and conspiracy theories. Recently, AI-generated deepfake videos, such as one of US congresswoman Alexandria Ocasio-Cortez, which depicted her in Congress arguing that a jeans ad featuring Sydney Sweeney was racist, have gone viral, highlighting new dangers in disinformation.
Kanisetti points out that many Indians consume historical materials now through videos. These don’t claim authenticity, but to the untrained mind the realistic-looking videos appear to be well-researched. “The right wing has no incentive to doctor primary sources yet, because the general public is already uninterested in the ambiguities of evidence-based history,” he said. Many viewers want affirmation of their beliefs; they no longer care if what they are seeing is accurate.
In What is History?, EH Carr wrote: “The facts speak only when the historian calls on them: it is he who decides to which facts to give the floor, and in what order or context.” The selection of facts is integral to historical research. Good historians must be aware of their personal biases and the context of their work. Only humans have the capability to understand the emotions involved and the ethical choices that need to be made; it is why at universities, history is part of the humanities department and not in a scientific lab.
Historians are quick to caution. One of the biggest risks is Holocaust denial and revisionism. Unesco published a report last year with the World Jewish Congress, which showed how hate groups could use AI to deny the Holocaust, including by fabricated testimonies and altered historical records. The Historical Figures app allows users to ‘chat’ with prominent Nazis such as Adolf Hitler and Joseph Goebbels, and it falsely claims that Goebbels (for example) was not intentionally involved in the Holocaust and had tried to prevent violence against Jews. Unesco has its own recommendations for ethics in AI, which it urges governments to follow. It has also asked tech companies to improve their standards and act responsibly.
But governments have neither the capacity nor, in some cases, the willingness to prevent disinformation from spreading widely. Can states be trusted? Can corporations, which are incentivised by maximising shareholder value? Open-source algorithms, digital watermarking and community-based content moderation are all necessary potential solutions, but none is sufficient. Educating the public, particularly young people, on how to critically evaluate information and recognise misinformation is crucial to combat the negative impacts.
Thum told me: “The only defence against this is by educating ourselves to be more sceptical and information-literate; by democratising the tools and skills of history; and by teaching people to be more sceptical and critical of those with information and power, so that we are less likely to be tricked by AI or whatever the next technology is.”
In 1987, I interviewed Salman Rushdie in Bombay, as the city was then known, when he had completed writing The Satanic Verses. Other than his editors, few knew what the novel was about.
When I asked him, he said: “It is about angels and devils and how it is very difficult to establish ideas of morality in a world which has become so uncertain that it is difficult to even agree on what is happening. When one can’t agree on a description of reality, it is very hard to agree on whether that reality is good or evil, right or wrong. When one can’t say what is actually the case, it is difficult to proceed from that to an ethical position. Angels and devils are becoming confused ideas … It is an attempt to come to grips with that sense of a crumbling moral fabric or, at least, a need for the reconstruction of old simplicities.”
AI makes us believe in simplicities, that truth is easy to access, and that answers to complex questions are just one click away. It is the product in our age of instant gratification. Reality, like history, is more nuanced. The challenge lies in our not being condemned to repeat it.
