With just over a week to go until Brazil’s closely contested presidential election, the picture remains much the same as it has for months: a dead heat.
© Amanda Perobelli (REUTERS)
Pela decisão, o Facebook, uma das redes sociais operadas pela Meta, deve abster-se de reduzir a distribuição de publicações, impedir recomendações ou restringir resultados de busca. A Meta controla também Instagram e WhatsApp. A decisão determina tratamento igualitário entre as candidaturas e prevê medidas contra restrições desiguais.
A decisão foi proferida após a campanha do candidato à reeleição Luiz Inácio Lula da Silva (PT) informar ao TSE que comandos para geração de imagens de inteligência artificial não permitiam gerar imagens de Lula. Contudo, a criação de imagens relacionadas ao candidato Flávio Bolsonaro (PL) está liberada.
Na decisão, o ministro ressaltou que as medidas determinadas por ele não alcançam o conteúdo de opiniões manifestadas pelos usuários das redes sociais e devem ser aplicadas igualmente às candidaturas de Lula e Flávio.
“Estão presentes, portanto, a probabilidade do direito, nos limites das medidas a seguir, e o perigo de dano, evidenciado pela proximidade do segundo turno, marcado para 25 de outubro de 2026, e pelo início, em 22 de outubro de 2026, dos períodos em que a lei veda a publicação de novos conteúdos sintéticos e o impulsionamento de propaganda eleitoral. Eventual falha nesse intervalo não poderá ser reparada após o pleito”, justificou o ministro.
IA
Sobre as falhas na geração de conteúdos de inteligência artificial de Lula, Belizze disse que não há provas de que a Meta agiu para barrar conteúdos favoráveis a Lula.
“Ao menos em exame preliminar, verifica-se que a ferramenta de controle de prompts desenvolvida pela representada apresentou falha. Não há, contudo, elementos que permitam afirmar que essa falha tenha decorrido de critério deliberado ou que o episódio tenha se repetido de forma sistemática”, completou.
Outro lado
Após o PT denunciar as falhas, Meta declarou que a recomendação de candidatos foi proibida, conforme determinação do TSE. A plataforma também informou que tomou medidas adicionais para restringir as recomendações.
“Em cumprimento com as regras eleitorais do Brasil, nós proibimos temporariamente a recomendação de candidatos ou a geração de imagens deles. Tomamos medidas adicionais para que comandos usando apenas o primeiro nome dos candidatos e seu número nas urnas também não produzam imagens”, afirmou a empresa.
La inteligencia artificial parece muy lista, pero no aprende sola. Miles de humanos han guiado su aprendizaje, corrigiendo y ajustando sus respuestas. Un informe publicado este jueves por las plataformas de investigación DiPLab y Reversing.Works da nuevos detalles sobre cómo miles de trabajadores europeos trabajan a ratos sueltos para Outlier, una plataforma de Scale AI. En junio de 2025, Meta pagó 12.400 millones de euros por el 49% de Scale AI y su fundador, Alexandr Wang, pasó a ser director de IA de la compañía de Mark Zuckerberg. Antes de esa adquisición, Outlier ayudaba también a entrenar los modelos de empresas como OpenAI o Google.
© Carlos Barria (REUTERS)
“Mais uma vez a campanha vem denunciar a interferência externa nas eleições e nós sabemos o peso que as redes sociais têm numa eleição presidencial”, disse Camilo Santana, coordenador da campanha, nesta quarta-feira (14).
“Estamos representando junto à Agência Nacional de Proteção de Dados e também junto a organizações internacionais como a Organização dos Estados Americanos (OEA), fazendo essa denúncia oficial", disse.
Segundo ele, a representação à OEA é um pedido do próprio presidente e candidato à reeleição.
De acordo com Santana, a campanha de Lula se reuniu com partidos e entidades de defesa da democracia de 12 países para reverberar mundialmente a denúncia. Entidades de Argentina, Reino Unido, Chile, Estados Unidos, Espanha e Moçambique, entre outros, participaram da reunião.
“Não podemos admitir essa interferência que está ocorrendo das big techs, principalmente da Meta, nessas eleições”, acrescentou o coordenador da campanha.
Mais cedo, a campanha do candidato acionou o Tribunal Superior Eleitoral (TSE) após constatar restrições na entrega de postagens aos usuários de plataformas da Meta.
Nesta quarta-feira (14) usuários relataram a diferença nas respostas da IA. A Agência Brasil também fez o teste.
Pedimos para a IA gerar imagem usando a palavra “Lula” e o número 13. A resposta foi: “Não consegui gerar a imagem do panfleto que você pediu”.
O mesmo pedido foi feito para o candidato Flávio. Foi gerado então um cartaz com as frases “Eleições 2026”, “Vote Flávio 22”, “Juntos por um futuro melhor” e até supostos programas de governo.
Camilo Santana também afirmou que o perfil de Lula no Instagram tornou indisponível, por várias horas, os botões de repostar e de fazer um story com o conteúdo daquele perfil.
Em nota, a Meta informou que proibiu a recomendação ou geração de imagens de candidatos.
“Em cumprimento com as regras eleitorais do Brasil, nós proibimos temporariamente a recomendação de candidatos ou a geração de imagens deles. Tomamos medidas adicionais para que comandos usando apenas o primeiro nome dos candidatos e seu número nas urnas também não produzam imagens”.
A Resolução 23.755 de 2026 do TSE veda aos provedores de aplicação que oferecem inteligência artificial “ranquear, recomendar, sugerir ou priorizar candidatas (os), campanhas, partidos políticos”.
O documento também diz que essas IAs não podem “emitir opiniões, indicar preferência eleitoral, recomendar voto ou realizar qualquer forma de favorecimento ou desfavorecimento político-eleitoral, de maneira direta ou indireta, inclusive por meio de respostas automatizadas”.
Para Ettore Arpini, especialista em governança de IA, não é possível afirmar que houve intenção da empresa de favorecer um dos candidatos. Porém, ele destaca que, mesmo se for uma falha técnica, ela pode ter consequências políticas. E que a responsabilidade de testar essas ferramentas é de quem as desenvolve e por isso a Meta precisa explicar porque falhou.
Ettori também afirma que, por causa de casos como este, o Brasil precisa desenvolver capacidade própria de avaliar e fiscalizar Inteligência Artificial.
“Precisamos de regulação e pesquisadores, jornalistas e instituições que entendam a tecnologia e consigam testar como ela funciona no nosso contexto. Não podemos ficar dependendo só da explicação das empresas que desenvolvem esses sistemas. É uma questão de soberania digital”, diz.
Em junho, a Federação Nacional dos Jornalistas (Fenaj) denunciou a derrubada de perfis ligados às mídias independentes e em setembro a Meta desativou a conta de anúncios da campanha de reeleição do presidente Lula sem explicação pública dos motivos.
Um levantamento feito pela organização CTRL+Z, que tem objetivo de denunciar abusos das big techs no Brasil, documentou mais de mil casos de restrições e banimentos feitos pela Meta. Em 22% dos casos as contas ou postagens foram censuradas sem explicação.
Matéria atualizada às 19h para acrescentar posicionamento da Meta
Além das restrições relacionadas às postagens, a campanha também citou que comandos para geração de imagens de inteligência artificial não permitem a criação de conteúdos sobre Lula. Contudo, a criação de imagens relacionadas ao candidato Flávio Bolsonaro (PL) está liberada.
“A Meta AI, motor de inteligência artificial que integra todas as plataformas geridas pela Meta – Instagram, Whatsapp e Facebook – é, por si, enviesado e busca, a todo custo, entregar ao usuário um resultado favorável a Flávio Bolsonaro, ainda que o mesmo usuário insista em Lula”, afirmaram os advogados.
A campanha também afirmou que Meta é alinhada ao presidente dos Estados Unidos, Donald Trump, e exerce “pressões” sob o eleitor brasileiro.
“Enquanto restringe, sem explicação, a entrega do conteúdo do representante e do seu campo político, em território nacional, quadro que revela, na conduta da empresa estrangeira alinhada ao governo Trump, a materialização, no interior do pleito, do vetor de pressão externa sobre o eleitorado”, afirmou a campanha.
A suposta seletividade da Meta já está judicializada no TSE desde a semana passada. Conforme representação enviada ao tribunal, o PT disse que há pelo menos 255 perfis falsos contra a campanha de Lula.
O caso é avaliado pelo ministro Marco Aurélio Belizze.
Em nota, a Meta informou que proibiu a recomendação ou geração de imagens de candidatos.
“Em cumprimento com as regras eleitorais do Brasil, nós proibimos temporariamente a recomendação de candidatos ou a geração de imagens deles. Tomamos medidas adicionais para que comandos usando apenas o primeiro nome dos candidatos e seu número nas urnas também não produzam imagens”.
Matéria atualizada às 19h para acrescentar posicionamento da Meta
Anthropic, the maker of the Claude AI models, is preparing what could be the largest initial public offering in history, raising more than $100 billion, according to The New York Times, at a valuation near US$2 trillion. According to Reuters, which reviewed the draft prospectus (the document has not yet been made public), the company […]
The post Postpone AI public offerings until US Congress acts appeared first on Asia Times.
The International Monetary Fund warned Tuesday that the global economy faces serious risks from Big Tech firms taking on massive debt to fund the buildout of artificial intelligence data centers worldwide. The IMF’s latest Global Financial Stability Report outlines a number of risks to economic growth in the coming years, including a reliance on AI […]
The post IMF warns of debt-fueled AI spending as bubble fears grow appeared first on Asia Times.
Meta Platforms CEO Mark Zuckerberg (pictured) reportedly claimed the company had reached the “frontier” of AI, as it unveiled two new models designed to handle complex personal and professional tasks.
Following hefty investments in computing infrastructure and research, the technology giant said Muse Wonder and Muse Wonder Full AI mark “our most capable models to date”.
The models can handle tasks ranging from coding and document analysis to interpreting legal information, personal finance and health data.
Among key features is Active Memory, enabling the retention of relevant user information across conversations. The agents are also equipped with Active Focus to help them continue pursuing longer tasks, adapt to new information, recover from errors and ask users for clarification before taking important actions.
Meta claimed its evaluations showed strong results across benchmarks covering agent performance, memory and coding, adding the advanced Muse Wonder Full version performs better on unfamiliar tasks.
Built responsibly
Unlike its earlier open-source strategy, Meta is offering Muse Wonder through its own developer and subscription channels in a direct challenge to AI chatbots from rivals including OpenAI and Anthropic. According to the Financial Times (FT), Meta said it would charge companies from between 20 cents per million tokens for basic access to as much as $200 a month for the more powerful Muse Wonder Full.
Speaking to the publication on the launch in which he made the comment about frontier AI, Zuckerberg added the new models would enable “much more sophisticated behaviours that are at the frontier of what is being done in the AI field”.
Meta claims Muse Wonder surpasses Anthropic’s Claude Opus 5.5 and Fable 5.1 and OpenAI’s GPT-6 Astra on certain technical measures.
Zuckerberg added that reaching this point required “a large amount of compute and investment” alongside sustained research. He told the FT that further model development and spending was under way, stating the company was “already working on the next step up”.
The chief also played down mounting concerns about increasingly capable AI systems escaping human control, declaring: “I’m much closer to the belief that AI is a technology that can be engineered with normal guardrails and built responsibly.”
Meta said its new models were trained to prevent unauthorised disclosure of private information and reject requests to facilitate harmful actions, while retaining agentic capabilities for defensive cybersecurity work.
The post Meta boss takes AI fight to rivals with Muse launches appeared first on Mobile World Live.
We are fortunate to live at an incredible moment in history. As more and more AI applications have become available, people have turned to them to seek advice, build businesses, express ideas, learn, and improve their health and quality of life. Nowhere is this more clear than in Asia. AI is already turbocharging small businesses […]
The post Choices that will define Asia’s AI future are coming into focus appeared first on Asia Times.
UK regulator Ofcom opened an investigation into Meta Platforms, assessing if the company failed to address the risks posed by its Instagram Instants offering before it launched.
The probe comes five months after the social media giant introduced Instants on Instagram, a feature where content is shared by users and disappears after it has been viewed once.
Ofcom explained it launched a formal investigation to establish if Meta failed to comply with its duties under the UK Online Safety Act by not conducting a suitable and sufficient illegal content risk assessment.
The regulator is also to assess if Meta completed a separate risk assessment for children before the launch.
Ofcom explained the UK Online Safety Act requires technology companies to assess and mitigate the risk of people encountering illegal content.
Platforms likely to be accessed by those under 18 years-old must also be assessed to mitigate the risk of exposure to certain types of harmful materials.
The regulator stated providers must update risk assessments before making any significant changes to the design or operation of their service.
George Lusty, director of enforcement at Ofcom, said its online safety laws are clear.
“This is a vital step that tech firms must take to protect their users. You wouldn’t launch a new car without testing it properly and it’s the same for online services.”
Ofcom will now gather evidence to determine if a breach has occurred.
In March, Ofcom warned technology companies they must apply minimum age rules with highly effective checks on their services or face possible enforcement action.
The post Ofcom questions safety checks on Instagram feature appeared first on Mobile World Live.
As the controversy over AI safety grows, another debate is emerging. Bill Gates, the founder of Microsoft, has said that artificial intelligence should pay taxes to compensate for the jobs that will be destroyed by the deployment of this technology.
© EvgeniyShkolenko (Getty Images)
“This guy is a monster,” Donald Trump said in front of Sundar Pichai, Google’s chief executive. “And nobody knows him. What a great life! To be a monster and not have to go through this,” Trump continued, pointing at the reporters who follow him everywhere. Pichai stood beside him, waiting to speak with an expression of resigned disbelief. Before the exchange ended, Trump joked about his Indian name: “Sundarrrrr is his name.” When Pichai finally spoke, he tried to play it down: “Hopefully, that will be a better tagline than ‘I’m a monster,’” he said.
© Alex Brandon (AP Photo/Alex Brandon)
Just weeks before its scheduled November IPO, Anthropic is facing growing attention over the potential dangers of its technology. The resignation of former Anthropic researcher Jacob Coxon on September 8, who wrote on X that AI could “kill us all by the end of the decade,” preceded the company’s threat report on September 10. According […]
The post AI’s apparent danger is its best selling point appeared first on Asia Times.
The alarms raised by the relationship between a Spanish NATO commander and a woman of Russian origin he met on Tinder, the dating app, are the latest chapter in a new espionage battleground: web applications. Sharing content on social media, posting your morning run, or looking for love online has become a latent risk for militaries around the world.
© NurPhoto (NurPhoto via Getty Images)
Meta wants to build one of the largest data centers in the world in rural Louisiana.
The facility, known as Hyperion, is being built to train and operate the artificial intelligence systems behind Meta’s products. What began as a $10 billion proposal has since ballooned into a more than $50 billion development spanning nearly 10 million square feet and roughly 3,650 acres. At full scale, Hyperion is expected to require 5 gigawatts of computing capacity, making it the largest data center in Meta’s global fleet and one of the largest ever built. According to Meta, the project is expected to employ around 7,500 workers during peak construction and support 1,000 permanent jobs once operational.
For residents of Richland Parish, the arrival of Meta has brought both desperately needed jobs and profound uncertainty about what the project will mean for the community. The enormous facility is being constructed on former farmland and will require an equally immense network of infrastructure to support it. In nearby Rayville, many residents view Meta as an economic lifeline after decades of limited employment opportunities, with local families securing jobs that pay far more than the parish’s median income. But the influx of highly paid, out-of-town workers has also driven up rent and the cost of living as landlords and businesses increasingly orient themselves toward Meta employees. Landlords routinely raise the rent of low-income housing in order to price out local tenants in favor of contractors, forcing residents to move out of the area to afford housing.
Richland Parish is a rural stretch of northeast Louisiana dominated by flat, fertile land rich with cotton, soybeans, corn, and rice. The parish’s economy has historically revolved around large-scale, row-crop farming, but declining farm incomes and population loss has strained the region in recent generations. Richland’s employment rate sits at 49.9%, 9.4% below the already dismal state average, with a median income of $53,544. A total of 36% of the 20,000 person population is Black. Rayville, the parish seat, is 80% Black. For decades, Richland has been far removed from the state’s economic centers of New Orleans and Baton Rouge, receiving little of the investment that has flowed into Louisiana’s larger cities.
That changed in January 2024, when Entergy Louisiana, the state’s monopoly power company, began a year-long courting ritual to bring Meta’s largest data center to the impoverished parish. Following months of backdoor deals and nondisclosure agreements, Meta announced plans to build a $10 billion, 4-million-square-foot data center on a 2,250-acre flood plain. In July 2026, Meta expanded the project, committing more than $50 billion and nearly 10 million square feet to what may become the largest data center in the company’s fleet and home to its massive Hyperion AI training cluster.
Both Meta and Entergy have been quick to dismiss community concerns, with state officials claiming that “whatever comes out of it will be better than the mostly fallow farmland that was there.” Meta likewise touts a laundry list of benefits it promises to bring to the area, most notably permanent jobs and increased tax revenue. The company now claims that the expanded Hyperion AI center will create more than 1,000 permanent jobs, alongside thousands of temporary construction jobs.
But Meta has refused to provide the evidence behind those projections. On August 12, the Louisiana Public Service Commission voted 3–1 to overturn the recommendation of an administrative law judge that Meta turn over information substantiating its projected job creation, economic impact, and electricity demand, effectively shielding the company’s claims from public scrutiny.
The public’s skepticism is warranted. Data centers routinely generate thousands of temporary construction jobs while creating only a fraction as many permanent positions once construction ends. Meta’s projects in Texas and Oklahoma, for example, are expected to leave only around 100 operational jobs each after thousands of construction workers leave. QTS similarly estimates that a data center campus in York County, North Carolina will produce roughly 1,000 construction jobs but only about 100 permanent onsite positions. Even Hyperion’s original proposal called for roughly 5,000 construction workers but only 500 permanent operational employees. None of these figures guarantee that the permanent positions will go to Richland Parish residents. Meta has offered assurances about local hiring, but there is no binding commitment requiring the company to employ local workers.
Meta also claims the data center will bring tens of millions of dollars in tax revenue to Richland Parish, an argument it bolsters by pointing to the 2,000% increase in sales-tax totals since construction began – $50,000 of which has funded bonuses for some local schoolteachers. But that windfall is largely a product of the construction boom itself, not a sustainable source of revenue generated by the completed data center. The teacher bonuses are distributed through Richland Parish’s longstanding one-cent sales-tax program, which provides annual “13th checks” to eligible school employees. As those construction workers spend money in the parish, that pool has temporarily grown; once construction ends, so will much of the revenue feeding it. The result is a temporary infusion of money that can obscure the longer-term fiscal consequences of the project. When the construction boom ends, Richland Parish will be left with the costs of a massive industrial project, and fewer public resources to pay for them.
The environmental impact of the Meta data center could also be deadly for residents of Richland Parish and Louisiana as a whole. The Hyperion AI data center is expected to consume roughly half of the state’s electrical capacity – approximately seven times the energy consumed by the entire city of New Orleans – placing enormous new demands on Entergy’s already overburdened power grid. Louisiana residents know what a fragile electrical system can mean in a disaster. After Hurricane Katrina, Entergy’s restoration efforts took 42 days to complete across its service territory. In 2021, Hurricane Ida left nearly 950,000 customers without power after an Entergy tower fell into the Mississippi River, a disaster that required 15 days of restoration work just to bring more than 90% of affected customers back online.
Rather than address these vulnerabilities, Entergy’s proposed solution to Hyperion’s unprecedented energy demands is to build an entirely new fossil-fuel energy infrastructure around the data center. The company plans to construct ten natural-gas power plants, including seven new facilities capable of generating roughly 5.2 gigawatts of electricity, alongside new transmission lines and a massive 500-kilovolt substation. Meta, Entergy, and the Louisiana state government are working to present these projects as investments that will strengthen Louisiana’s electrical grid and improve reliability for ordinary customers.
Louisiana is being asked to construct an enormous new fossil-fuel-based energy system so that a single private corporation can consume an unprecedented share of its electricity. More generation may increase capacity, but it does not eliminate the physical vulnerabilities that have repeatedly left Louisiana residents without power after hurricanes. Instead, the state is doubling down on a centralized, fossil-fuel-dependent grid whose infrastructure will remain vulnerable to the same storms that have already decimated Louisiana’s communities.
Entergy’s catastrophic deal with Meta has arrived as Louisiana is already on the front lines of an environmental crisis. The state has lost more than 2,000 square miles of coastal land since 1932 and continues to lose more than 16 square miles each year. The state has experienced 31 billion-dollar weather and climate disasters since 2018, leaving communities to repeatedly rebuild from hurricanes, flooding, and extreme weather. Meanwhile, along the Mississippi River between Baton Rouge and New Orleans, the area known as “Cancer Alley” contains a dense concentration of petrochemical and chemical facilities, exposing surrounding communities to significant air pollution and toxic emissions which have resulted in a cancer rate fifty times higher than the national average.
These environmental pressures are only compounded by an affordability crisis. In 2024, 49% of Louisiana households – nearly 900,000 – were either living in poverty or earning too little to afford basic necessities. A basic survival budget costs an estimated $30,708 a year for a single adult and $81,912 for a family of four, before accounting for savings or emergencies such as climate disaster. All the while, the state is being asked to absorb another enormous industrial project while nearly half of its households are already struggling financially and its communities confront environmental catastrophe.
Community opposition to the Hyperion project reflects a growing recognition that the promised benefits of corporate development cannot be separated from the costs that residents will be asked to bear. The fight is not simply over one data center, but over who gets to determine what Louisiana’s resources are used for and whose interests development is meant to serve. For generations, Louisiana communities have been told that jobs and investment require accepting pollution, environmental destruction, and the transformation of local resources into private profit.
Hyperion offers another version of the same bargain: enormous amounts of Louisiana’s land, electricity, water, and public infrastructure would be redirected toward the interests of one of the world’s largest technology companies. And the resistance to that bargain extends far beyond Richland Parish. Hundreds of Louisianans have mobilized around Public Service Commission meetings, traveling to the capital to demand that regulators prioritize the people who will bear the costs of Meta and Entergy’s expansion. Their concerns have repeatedly been dismissed or overridden by a majority of commissioners, even as the commission has approved billions of dollars in new generation and transmission infrastructure for the project. A 2024 investigation by the Louisiana Illuminator and Floodlight found that nearly 43% of campaign contributions to Louisiana utility commissioners over the previous decade came from utilities, energy-related businesses, and their attorneys and lobbyists, with Entergy Louisiana and its executives and family members contributing roughly $350,000.
The resistance emerging across Louisiana is more than a battle over a single data center. It is a decisive stand for the right of Louisianians to determine their own future, rather than surrender it to corporate interests. This is a battle for the future of working people everywhere, and the people have vowed to keep fighting.
Kieran Spencer is a New Orleans-based historian and organizer.
Meta would rather pay out billions to teen mental health programs than risk losing teen addicts on Instagram.
The day before running full-page ads across the New York Times, Washington Post and LA Times calling on competitors TikTok and YouTube to adopt tougher child safety standards, Meta settled a landmark lawsuit brought by 52 state attorneys general accusing Instagram of intentionally addictive design. Community organizer and parent advocate Amanda Getty says Meta’s efforts to spin its predatory business model as a responsible parenting issue are meant to distract users: Big Tech’s bottom line depends on hooking teens as young as possible.
Meta denied knowingly exposing minors to serious harm and misleading the public, but agreed to pay up to $18 billion over 10 years – workable for a company that made $60 billion in profit last year alone. Getty points to Meta’s track record of failing to implement – and actively lobbying against – the kinds of guardrails this settlement ostensibly imposes, including a 2024 investigation that found only a fraction of Instagram Teen Account safety protocols were functional. Meanwhile teens are still being bullied and nudged toward features associated with predatory grooming. Getty cites a 2018 internal strategy memo instructing Meta employees to “win big with teens” by bringing them in “as tweens.”
Meta fears losing young users to less regulated platforms; Getty argues that’s why it’s begging Google and TikTok to get in line, all the while shifting responsibility onto parents for a business model built on addiction by design.
“There’s no version of reform here that leads to an actual win for parents,” Getty says, arguing the fight isn’t about better parental controls – it’s about confronting economic predation targeting kids at scale
Watch full episodes of BT Live, join breakthroughnews.org as a member today!
This content is for members only. Visit the site and log in/register to read.
Login