An anonymous reader quotes a report from the Associated Press: They're cute, even cuddly, and promise learning and companionship -- but artificial intelligence toys are not safe for kids, according to children's and consumer advocacy groups urging parents not to buy them during the holiday season. These toys, marketed to kids as young as 2 years old, are generally powered by AI models that have already been shown to harm children and teenagers, such as OpenAI's ChatGPT, according to an advisory published Thursday by the children's advocacy group Fairplay and signed by more than 150 organizations and individual experts such as child psychiatrists and educators.

"The serious harms that AI chatbots have inflicted on children are well-documented, including fostering obsessive use, having explicit sexual conversations, and encouraging unsafe behaviors, violence against others, and self-harm," Fairplay said. AI toys, made by companies including Curio Interactive and Keyi Technologies, are often marketed as educational, but Fairplay says they can displace important creative and learning activities. They promise friendship but disrupt children's relationships and resilience, the group said. "What's different about young children is that their brains are being wired for the first time and developmentally it is natural for them to be trustful, for them to seek relationships with kind and friendly characters," said Rachel Franz, director of Fairplay's Young Children Thrive Offline Program. Because of this, she added, the trust young children are placing in these toys can exacerbate the types of harms older children are already experiencing with AI chatbots.

A separate report Thursday by Common Sense Media and psychiatrists at Stanford University's medical school warned teenagers against using popular AI chatbots as therapists. Fairplay, a 25-year-old organization formerly known as the Campaign for a Commercial-Free Childhood, has been warning about AI toys for years. They just weren't as advanced as they are today. A decade ago, during an emerging fad of internet-connected toys and AI speech recognition, the group helped lead a backlash against Mattel's talking Hello Barbie doll that it said was recording and analyzing children's conversations. This time, though AI toys are mostly sold online and more popular in Asia than elsewhere, Franz said some have started to appear on store shelves in the U.S. and more could be on the way. "Everything has been released with no regulation and no research, so it gives us extra pause when all of a sudden we see more and more manufacturers, including Mattel, who recently partnered with OpenAI, potentially putting out these products," Franz said. Last week, consumer advocates at U.S. PIRG called out the trend of buying AI toys in its annual "Trouble in Toyland" report. This year, the organization tested four toys that use AI chatbots. "We found some of these toys will talk in-depth about sexually explicit topics, will offer advice on where a child can find matches or knives, act dismayed when you say you have to leave, and have limited or no parental controls," the report said.

An anonymous reader quotes a report from Bloomberg: The world's largest music companies have licensed their works to a music startup called Klay, which is building a streaming service that will allow users to remake songs using artificial intelligence tools. Klay is the first music AI service to reach a deal with all three major record labels, Universal Music Group NV, Sony Music and Warner Music Group Corp., according to people familiar with the deals. Klay plans to announce its agreements in the coming days, said the people, who asked not to be identified discussing confidential plans.

Klay is building a product that will offer the features of a streaming service like Spotify, amplified by AI technology that will let users remake songs in different styles. Klay has licensed the rights to thousands of hit songs so that it can train its large language model. The company has positioned itself as a friend of the industry, offering assurances that the artists and labels will have some control over how their work is used. Klay is led by music producer Ary Attie and also employs former executives from Sony Music and Google's DeepMind, an AI laboratory.

An anonymous reader quotes a report from Ars Technica: Google's meme-friendly Nano Banana image-generation model is getting an upgrade. The new Nano Banana Pro is rolling out with improved reasoning and instruction following, giving users the ability to create more accurate images with legible text and make precise edits to existing images. It's available to everyone in the Gemini app, but free users will find themselves up against the usage limits pretty quickly. Nano Banana Pro is part of the newly launched Gemini 3 Pro -- it's actually called Gemini 3 Pro Image in the same way the original is Gemini 2.5 Flash Image, but Google is sticking with the meme-y name. You can access it by selecting Gemini 3 Pro and then turning on the "Create images" option.

Google says the new model can follow complex prompts to create more accurate images. The model is apparently so capable that it can generate an entire usable infographic in a single shot with no weird AI squiggles in place of words. Nano Banana Pro is also better at maintaining consistency in images. You can blend up to 14 images with this tool, and it can maintain the appearance of up to five people in outputs. Google also promises better editing. You can refine your AI images or provide Nano Banana Pro with a photo and make localized edits without as many AI glitches. It can even change core elements of the image like camera angles, color grading, and lighting without altering other elements. Google is pushing the professional use angle with its new model, which has much-improved resolution options. Your creations in Nano Banana Pro can be rendered at up to 4K.

Microsoft's push to transform Windows into an "agentic OS" that allows AI agents to control PCs is drawing user backlash similar to the Windows 8 controversy, as the company marks the operating system's 40th anniversary this week, writes Tom Warren, a reporter at The Verge who has been covering Microsoft for nearly two decades. Windows chief Pavan Davuluri announced the agentic OS plans in a post on X last week and faced immediate criticism in hundreds of replies before they were locked days later.

"It's evolving into a product that's driving people to Mac and Linux," one person wrote, while another asked for a return to Windows 7's "clean UI, clean icon, a unified control panel, no bloat apps, no ads, just a pure performant OS." Davuluri later responded to software engineer Gergely Orosz, saying "we care deeply about developers" and acknowledging Microsoft has "work to do on the experience, both on the everyday usability, from inconsistent dialogs to power user experiences."

Microsoft CEO Satya Nadella told the Dwarkesh Podcast that the company's business "which today is an end user tools business, will become, essentially an infrastructure business in support of agents doing work." The Recall feature already spooked users when it was initially turned on by default before Microsoft reworked it to be opt-in. Navjot Virk, corporate vice president of Windows experiences, told The Verge that "every user can use [AI agents] when they're ready. It's their choice, they decide."

Autonomous electric tractor startup Monarch Tractor -- which we covered in 2022 -- warned staff Thursday it may need to lay off more than 100 employees, or possibly even "shut down," according to a company-wide memo obtained by TechCrunch. The report adds: The memo comes after Monarch Tractor was already cutting some positions over the last few weeks at its California corporate facilities and remote teams in India and Singapore, according to multiple former employees who spoke with TechCrunch on the condition of anonymity.

Monarch Tractor was founded in 2018 by a team that included a former top executive at Tesla's first gigafactory and Carlo Mondavi, a scion of the famous winemaking family. The company raised at least $220 million, including $133 million in 2024, as it pursued a goal of making "driver optional" autonomous tractors that could perform tasks at places like wineries and other fruit farms.

An anonymous reader shares a report: A Microsoft executive is questioning why more people aren't impressed with AI, a week after the company touted the evolution of Windows into an "agentic OS," which immediately triggered backlash.

"Jeez there so many cynics! It cracks me up when I hear people call AI underwhelming," tweeted Mustafa Suleyman, the CEO for Microsoft's AI group. Suleyman added that he grew up playing the old-school 2D Snake game on a Nokia phone. "The fact that people are unimpressed that we can have a fluent conversation with a super smart AI that can generate any image/video is mindblowing to me," he wrote.

OpenAI won't say who audits its books. The company, which projects to hit an ARR of $20 billion this year and is valued at $500 billion, has committed to spending about $1.4 trillion on data centers over the next decade. It accounts for roughly two-thirds of unfulfilled contracts at Oracle and two-fifths at CoreWeave. Microsoft alone holds around $375 billion in unfulfilled contracts with OpenAI.

Reuters reported the company may target a $1 trillion valuation for a potential IPO in coming years. Most companies at this scale use one of the Big Four accounting firms: Deloitte, EY, KPMG or PwC. OpenAI declined to comment to Financial Times. A person close to the organization told the publication the company has "an industry standard audit with one of the Big Four firms." The company's latest Form 990 filing lists Fontanello, Duffield, & Otake -- a small San Francisco accountancy firm -- as the paid preparer. The form does say an independent accountant audited the statements.

Michael Burry, last night: "Can anyone name [OpenAI's] auditor?"

An anonymous reader quotes a report from Politico: The White House is preparing to issue an executive order as soon as Friday that tells the Department of Justice and other federal agencies to prevent states from regulating artificial intelligence, according to four people familiar with the matter and a leaked draft of the order obtained by POLITICO. The draft document, confirmed as authentic by three people familiar with the matter, would create an "AI Litigation Task Force" at the DOJ whose "sole responsibility" would be to challenge state AI laws.

Government lawyers would be directed to challenge state laws on the grounds that they unconstitutionally regulate interstate commerce, are preempted by existing federal regulations or otherwise at the attorney general's discretion. The task force would consult with administration officials, including the special adviser for AI and crypto -- a role currently occupied by tech investor David Sacks.

The executive order, in the draft obtained by POLITICO, would also empower Commerce Secretary Howard Lutnick to publish a review of "onerous" state AI laws within 90 days and restrict federal broadband funds to states whose AI laws are found to be objectionable. It would direct the Federal Trade Commission to investigate whether state AI laws that "require alterations to the truthful outputs of AI models" are blocked by the FTC Act. And it would order the Federal Communications Commission to begin work on a reporting and disclosure standard for AI models that would preempt conflicting state laws.

Canadian librarian Ian Linkletter has ended a five-year legal battle with ed-tech firm Proctorio after being sued for sharing public YouTube help videos that exposed how the company's remote-proctoring AI works. Ars Technica reports: ... Together, the videos, the help center screenshot, and another screenshot showing course material describing how Proctorio works were enough for Proctorio to take Linkletter to court. The ed tech company promptly filed a lawsuit and obtained a temporary injunction by spuriously claiming that Linkletter shared private YouTube videos containing confidential information. Because the YouTube videos -- which were public but "unlisted" when Linkletter shared them -- had been removed, Linkletter did not have to delete the seven tweets that initially caught Proctorio's attention, but the injunction required that he remove two tweets, including the screenshots.

In the five years since, the legal fight dragged on, with no end in sight until last week, as Canadian courts tangled with copyright allegations that tested a recently passed law intended to shield Canadian rights to free expression, the Protection of Public Participation Act. To fund his defense, Linkletter said in a blog announcing the settlement that he invested his life savings "ten times over." Additionally, about 900 GoFundMe supporters and thousands of members of the Association of Administrative and Professional Staff at UBC contributed tens of thousands more. For the last year of the battle, a law firm, Norton Rose Fulbright, agreed to represent him on a pro bono basis, which Linkletter said âoewas a huge relief to me, as it meant I could defend myself all the way if Proctorio chose to proceed with the litigation."

The terms of the settlement remain confidential, but both Linkletter and Proctorio confirmed that no money was exchanged. For Proctorio, the settlement made permanent the injunction that restricted Linkletter from posting the company's help center or instructional materials. But it doesn't stop Linkletter from remaining the company's biggest critic, as "there are no other restrictions on my freedom of expression," Linkletter's blog noted. "I've won my life back!" Linkletter wrote, while reassuring his supporters that he's "fine" with how things ended. "It doesn't take much imagination to understand why Proctorio is a nightmare for students," Linkletter wrote. "I can say everything that matters about Proctorio using public information."

An anonymous reader quotes a report from the New York Times: When Mark Zuckerberg, Meta's chief executive, unveiled the company's Superintelligence Lab in June, he named 11 artificial intelligence researchers who were joining his ambitious effort to build a machine more powerful than the human brain. All 11 were immigrants educated in other countries. Seven were born in China, according to a memo viewed by The New York Times. Although many American executives, government officials and pundits have spent months painting China as the enemy of America's rapid push into A.I., much of the groundbreaking research emerging from the United States is driven by Chinese talent.

Two new studies show that researchers born and educated in China have for years played major roles inside leading U.S. artificial intelligence labs. They also continue to drive important A.I. research in industry and academia, despite the Trump administration's crackdown on immigration and growing anti-China sentiment in Silicon Valley. The research, from two organizations, provides a detailed look at how much the American tech industry continues to rely on engineers from China, particularly in A.I. The findings also offer a more nuanced understanding of how researchers in the two countries continue to collaborate, despite increasingly heated language from Washington and Beijing.

Adobe is buying Semrush for $1.9 billion in a move to supercharge its AI-driven marketing stack. Reuters reports: Semrush designs and develops AI software that helps companies with search engine optimization, social media and digital advertising. The acquisition, expected to close in the first half of next year, would allow Adobe to help marketers better understand how their brands are viewed by online consumers through searches on websites and generative AI bots such as ChatGPT and Gemini. "The price is steep as Semrush isn't a massive revenue engine on its own, so Adobe is likely paying for strategic value. The payoff could be high too if Adobe can quickly turn Semrush's data into monetizable AI products," said Emarketer analyst Grace Harmon.

"While we are positive on Adobe restarting its M&A engine given the success that it has seen with this motion over the years... this deal likely does little to answer the questions revolving around the company's creative cloud business," added William Blair analysts.

pbahra writes: Saudi Arabia is betting that the future of Hollywood won't be built in physical stages but in datacenters. In a push to anchor itself in next-generation film production, Riyadh-based Humain has led Luma AI's latest Series C round, backing the shift towards cloud-based, AI-generated video rather than traditional studio infrastructure.. Humain's announcement says the new investment will accelerate Luma's development of world models capable of learning from video, audio and language to generate photorealistic scenes, environments and characters on demand. Supporters argue this could upend film-making by pushing much of Hollywood's production pipeline into high-performance datacenters rather than physical sets.

Nvidia blew past earnings expectations with soaring revenue and profit, easing fears of an AI bubble and reinforcing its position as the engine of the global AI boom. From a report: Nvidia's sales grew 62% year-over-year to $57 billion in the October quarter, ahead of the $54.9 billion Wall Street had projected, signaling that demand for AI chips remains strong even as more questions emerge about whether returns from the technology will keep up with the pace of AI infrastructure investments. It posted profits of $31.9 billion, up 65% from the year-ago quarter and also slightly above expectations.

"Blackwell sales are off the charts, and cloud GPUs are sold out," Nvidia CEO Jensen Huang said in a statement, a message that echoes his earlier arguments that fears of an AI bubble are overblown. The company also posted stronger-than-expected sales guidance of around $65 billion for the fourth quarter, another indicator that the AI spending spree isn't slowing anytime soon.

Linus Torvalds is "fairly positive" about vibe coding as a way for people to get computers to do things they otherwise could not. The Linux kernel maintainer made the comments during an interview at the Linux Foundation Open Source Summit in Seoul earlier this month. But he cautioned that vibe coding would be a "horrible, horrible idea from a maintenance standpoint" for production code.

Torvalds told Dirk Hohndel, head of open source at Verizon, that computers have become more complicated than when he learned to code by typing in programs from computer magazines. He said vibe coding offers a path into computing for newcomers. The kernel maintainer is not using AI-assisted coding himself. He said his role has shifted from rejecting new ideas to sometimes pushing for them against opposition from longstanding maintainers who "kind of get stuck in a rut."

Rust is "actually becoming a real part of the kernel instead of being this experimental thing," he said. Torvalds said AI crawlers have been "very disruptive to a lot of our infrastructure" because they gather data from kernel.org source code. Kernel maintainers receive bugs and security notices that are "made up by people who misuse AI," though the problem is smaller than for other projects such as curl.

Tsinghua University collected 4,986 AI and machine learning patents between 2005 and the end of 2024. The Beijing institution has received more than 900 patents last year alone. The total exceeds the combined patent count from MIT, Stanford, Princeton and Harvard during the same period. China now accounts for more than half of all active patent families globally in AI and machine learning fields, according to data analytics service LexisNexis.

The university also has more AI research papers among the 100 most cited than any other school at last count. The US still holds the most influential AI patents and the top performing models. Harvard and MIT consistently rank ahead of Tsinghua in patent influence. American institutions produced 40 notable AI models in 2024 compared to 15 from Chinese organizations, according to Stanford's AI Index Report. China's share of the world's elite AI researchers -- the top 2% -- rose from 10% in 2019 to 26% in 2022. The US share fell from 35% to 28% during the same period, according to the Information Technology & Innovation Foundation.

An anonymous reader quotes a report from CNBC: The Trump administration will provide Constellation Energy with a $1 billion loan to restart the Crane Clean Energy Center nuclear plant in Pennsylvania, Department of Energy officials said Tuesday. Previously known as Three Mile Island Unit 1, the plant is expected to start generating power again in 2027. Constellation unveiled plans to rename and restart the reactor in Sept. 2024 through a power purchase agreement with Microsoft to support the tech company's data center demand in the region.

Three Mile Island Unit 1 ceased operations in 2019, one of a dozen reactors that closed in recent years as nuclear struggled to compete against cheap natural gas. It sits on the same site as Three Mile Island Unit 2, the reactor that partially melted down in 1979 in the worst nuclear accident in U.S. history. The loan would cover the majority to the project's estimated cost of $1.6 billion. The first advance to Constellation is expected in the first quarter of 2026, said Greg Beard, senior advisor to the Energy Department's Loan Programs Office, in a call with reporters. The loan comes with a guarantee from Constellation that it will protect taxpayer money, Beard said.

An anonymous reader quotes a report from ZDNet: At KubeCon North America 2025 in Atlanta, the Cloud Native Computing Foundation (CNCF)'s leaders predicted an enormous surge in cloud-native computing, driven by the explosive growth of AI inference workloads. How much growth? They're predicting hundreds of billions of dollars in spending over the next 18 months. [...] Where cloud-native computing and AI inference come together is when AI is no longer a separate track from cloud-native computing. Instead, AI workloads, particularly inference tasks, are fueling a new era where intelligent applications require scalable and reliable infrastructure. That era is unfolding because, said [CNCF Executive Director Jonathan Bryce], "AI is moving from a few 'Training supercomputers' to widespread 'Enterprise Inference.' This is fundamentally a cloud-native problem. You, the platform engineers, are the ones who will build the open-source platforms that unlock enterprise AI."

"Cloud native and AI-native development are merging, and it's really an incredible place we're in right now," said CNCF CTO Chris Aniszczyk. The data backs up this opinion. For example, Google has reported that its internal inference jobs have processed 1.33 quadrillion tokens per month recently, up from 980 trillion just months before. [...] Aniszczyk added that cloud-native projects, especially Kubernetes, are adapting to serve inference workloads at scale: "Kubernetes is obviously one of the leading examples as of the last release the dynamic resource allocation feature enables GPU and TPU hardware abstraction in a Kubernetes context." To better meet the demand, the CNCF announced the Certified Kubernetes AI Conformance Program, which aims to make AI workloads as portable and reliable as traditional cloud-native applications.

"As AI moves into production, teams need a consistent infrastructure they can rely on," Aniszczyk stated during his keynote. "This initiative will create shared guardrails to ensure AI workloads behave predictably across environments. It builds on the same community-driven standards process we've used with Kubernetes to help bring consistency as AI adoption scales." What all this effort means for business is that AI inference spending on cloud-native infrastructure and services will reach into the hundreds of billions within the next 18 months. That investment is because CNCF leaders predict that enterprises will race to stand up reliable, cost-effective AI services.

Klarna has claimed that AI-related savings have allowed the buy now, pay later company to increase staff salaries by nearly 60%, but hinted it could slash more jobs after nearly halving its workforce over the past three years. From a report: Chief executive Sebastian Siemiatkowski said headcount had dropped from 5,527 to 2,907 since 2022, mostly as a result of natural attrition, with departing staff replaced by technology rather than by new staff members.

The figures add to the impact of an internal artificial intelligence programme, which had steadily reduced its use of outsourced workers including those in customer service, with technology now carrying out the work of 853 full-time staff, up from 700 earlier this year. It meant the company, which was founded in Sweden in 2005, had managed to increase revenues by 108% while keeping operating costs flat. Siemiatkowski told analysts on an earnings call on Tuesday that it was "pretty remarkable, and unheard of as a number, among businesses."

Microsoft's Copilot AI assistant in Windows 11 fails to replicate the capabilities shown in the company's TV advertisements. The Verge tested Copilot Vision over a week using the same prompts featured in ads airing during NFL games. When asked to identify a HyperX QuadCast 2S microphone visible in a YouTube video -- a task successfully completed in Microsoft's ad -- Copilot gave multiple incorrect answers. The assistant identified the microphone as a first-generation HyperX QuadCast, then as a Shure SM7b on two other occasions. Copilot couldn't identify the Saturn V rocket from a PowerPoint presentation despite the words "Saturn V" appearing on screen. When asked about a cave image from Microsoft's ad, Copilot gave inconsistent responses.

About a third of the time it provided directions to find the photo in File Explorer. On two occasions it explained how to launch Google Chrome. Four times it offered advice about booking flights to Belize. The cave is Rio Secreto in Playa del Carmen, Mexico. Microsoft spokesperson Blake Manfre said "Copilot Actions on Windows, which can take actions on local files, is not yet available." He described it as "an opt-in experimental feature that will be coming soon to Windows Insiders in Copilot Labs, starting with a narrow set of use cases while we optimize model performance and learn." Copilot cannot toggle basic Windows settings like dark mode. When asked to analyze a benchmark table in Google Sheets, it "constantly misread clear-as-day scores both in the spreadsheet and in the on-page review."

A majority of global fund managers think companies are overinvesting, as market anxiety grows about the sustainability of the AI spending boom. From a report: A net 20 per cent of fund managers surveyed this month by Bank of America said companies were spending too much on their investments -- the first time this has been a majority view in data running back to 2005. "This jump is driven by concerns over the magnitude and financing of the AI capex boom," said BofA analysts.

The surge in investment to develop AI infrastructure has been a dominant theme in the record rally in US tech stocks this year -- with chipmaker Nvidia becoming the world's first $5tn company last month -- but growing concerns about the sustainability of this spending has caused a pullback on Wall Street in recent weeks.