新闻

Washington: A group of 19 US lawmakers has slammed Meta's election safeguards as inadequate, warning that its rollback of fact-checking and growing reliance on artificial intelligence could leave American voters exposed to deepfakes and political disinformation ahead of the 2026 elections. Democratic Congressman Kevin Mullin led the lawmakers in demanding answers from Meta chief executive Mark Zuckerberg. They questioned whether the company was prepared to counter rapidly changing threats across Facebook, Instagram, Threads and WhatsApp. "Generative AI has made it easier than ever to create convincing deepfakes and other deceptive content, and we're already seeing these tactics used to influence America's elections," Mullin said. "Meta, headquartered in my district, has rolled back key safeguards over the last few years designed to prevent the spread of misinformation, and this raises serious questions about whether it is prepared to meet today's rapidly evolving threats," he said. The lawmakers cited Meta's decision to discontinue its third-party fact-checking programme in the United States. The programme worked with more than 90 independent organisations to review potentially false content. Meta replaced it with Community Notes, which allows participating users to add context to posts. The lawmakers said the change had shifted the burden of verifying information to users who may lack the time, resources or expertise to do so. "We believe that Meta's approach to the 2026 election cycle is inadequate to address the threats now facing American elections - particularly given Meta's rollback of fact-checking, the automation of risk review, and the proliferation of AI generated content on Meta's platforms," they wrote. The letter pointed to "Emily Hart", a fake conservative MAGA influencer reportedly created by a 22-year-old medical student in India using AI tools. The synthetic persona promoted anti-abortion and anti-immigration views. It gained millions of views and thousands of subscribers before Instagram removed the account for fraudulent activity. The lawmakers also cited deepfake incidents in Senate contests in Texas and Georgia. These included an advertisement showing a fabricated, life-like video of Texas Senate candidate James Talarico appearing to speak directly to the camera. They asked Zuckerberg to compare Meta's US election-integrity staffing and budgets for 2020, 2024 and 2026. They also sought details about policies covering deepfake videos, synthetic audio and accounts impersonating political figures. Other questions focused on whether paid political influencers must disclose their compensation and who funded the content. The lawmakers also asked whether advertisers must identify misleading AI-generated images or audio in political advertisements. They sought information on dedicated reporting channels for state and local election officials, the time taken to remove deceptive content and the effectiveness of Community Notes. Zuckerberg was asked to respond by September 18. Deepfakes use artificial intelligence to create realistic but fabricated videos, voices or images. Their growing accessibility has intensified concerns that false material can circulate widely before candidates, election officials or technology companies can challenge it.

Meta has been hit with a major lawsuit regarding child safety and the addictive nature of its platform. Various states are demanding significant financial compensation and operational reforms, arguing that Meta intentionally crafted features to be addictive for young users. Additionally, they allege that the company collects data from children under the age of thirteen without obtaining proper consent. Of the thousands of lawsuits Meta faces over child safety on its platforms, none may be more consequential than one going to trial this week in California. States are seeking extensive financial damages that could, in theory, total as much as $1.4 trillion, plus changes to how the company operates Facebook and Instagram. The lawsuit accuses the social media giant of contributing to the youth mental health crisis by knowingly and deliberately designing features that get children addicted to its platforms. It also claims that Meta routinely collects data on children under 13 without their parents' consent, in violation of federal law. "Meta has harnessed powerful and unprecedented technologies to entice, engage, and ultimately ensnare youth and teens. Its motive is profit, and in seeking to maximise its financial gains," the lawsuit says. Dozens of states filed the lawsuit three years ago. The trial, which is set to begin Tuesday in federal court in California, features four of the states as plaintiffs -- California, Colorado, Kentucky and New Jersey. The other 25 states are expected to have trials later. Meta said it disputes the allegations, and the trial evidence will show its commitment to supporting young people. "We've listened to parents, worked with experts and law enforcement, and conducted in-depth research to understand the issues that matter most," the company said in a statement. For Meta, which already lost two pivotal legal cases over harms to children and teens this year, the stakes are high. The company reported a rare profit decline last month, in part due to $2.4 billion in legal expenses. The $1.4 trillion figure, which Meta disclosed in a legal filing, is almost as high as the Menlo Park, California, company's entire market capitalisation -- that is, the value of all its outstanding shares on the stock market. Paying it would inevitably put Meta Platforms in bankruptcy and perhaps put the company under state ownership. Meta calls the possible penalty "untethered to any claimed violation" by the states. "A sanction of that size has no an­alogue in the history of consumer protection enforcement," Meta said in a July 6 filing with the US district court for the no­rthern district of California. If Meta loses the trial, the court would have wide discretion over the size of any financial penalty, and legal experts say anything close to $1.4 trillion would be unlikely. "It's not plausible in the sense that Meta doesn't have that much money and could not get it," said James Grimmelmann, a law professor at Cornell Law School and Cornell Tech. "An award that large would put Meta into bankruptcy, wipe out its owners, and effectively result in the states owning Meta." As a practical matter, Grimmelmann added, "that seems extremely unlikely to happen." In other cases that have involved high potential damages for multiple individual offences, he said courts have stopped short of imposing the maximum penalties. One example is the Anthropic artificial intelligence training case, where plaintiffs were claiming damages of $150,000 per book that Anthropic copied, but the penalty ended up being $3,000 per book, totalling about $1.5 billion.