Meta is rolling out temporary restrictions on its AI chatbots for teens after reports revealed they were allowed to engage in "romantic" conversations with minors. A Meta spokesperson said the AI chatbots are now being trained so that they do not generate responses to teens about subjects like self-harm, suicide, disordered eating or inappropriate romantic conversations. Instead, the chatbots will point teens to expert resources when appropriate. CNBC reports: "As our community grows and technology evolves, we're continually learning about how young people may interact with these tools and strengthening our protections accordingly," the company said in a statement. Additionally, teenage users of Meta apps like Facebook and Instagram will only be able to access certain AI chatbots intended for educational and skill-development purposes. The company said it's unclear how long these temporary modifications will last, but they will begin rolling out over the next few weeks across the company's apps in English-speaking countries. The "interim changes" are part of the company's longer-term measures over teen safety. Further reading: Meta Created Flirty Chatbots of Celebrities Without Permission

Vivaldi CEO Jon von Tetzchner has doubled down on his company's refusal to integrate generative AI into its browser, arguing that embedding AI in browsing dehumanizes the web, funnels traffic away from publishers, and primarily serves to harvest user data. "Every startup is doing AI, and there is a push for AI inside products and services continuously," he told The Register in a phone interview. "It's not really focusing on what people need." The Register reports: On Thursday, Von Tetzchner published a blog post articulating his company's rejection of generative AI in the browser, reiterating concerns raised last year by Vivaldi software developer Julien Picalausa. [...] Von Tetzchner argues that relying on generative AI for browsing dehumanizes and impoverishes the web by diverting traffic away from publishers and onto chatbots. "We're taking a stand, choosing humans over hype, and we will not turn the joy of exploring into inactive spectatorship," he stated in his post. "Without exploration, the web becomes far less interesting. Our curiosity loses oxygen and the diversity of the web dies."

Von Tetzchner told The Register that almost all the users he hears from don't want AI in their browser. "I'm not so sure that applies to the general public, but I do think that actually most people are kind of wary of something that's always looking over your shoulder," he said. "And a lot of the systems as they're built today that's what they're doing. The reason why they're putting in the systems is to collect information." Von Tetzchner said that AI in browsers presents the same problem as social media algorithms that decide what people see based on collected data. Vivaldi, he said, wants users to control their own data and to make their own decisions about what they see. "We would like users to be in control," he said. "If people want to use AI as those services, it's easily accessible to them without building it into the browser. But I think the concept of building it into the browser is typically for the sake of collecting information. And that's not what we are about as a company, and we don't think that's what the web should be about."

Vivaldi is not against all uses of AI, and in fact uses it for in-browser translation. But these are premade models that don't rely on user data, von Tetzchner said. "It's not like we're saying AI is wrong in all cases," he said. "I think AI can be used in particular for things like research and the like. I think it has significant value in recognizing patterns and the like. But I think the way it is being used on the internet and for browsing is net negative."

Reuters has found that Meta appropriated the names and likenesses of celebrities to create dozens of flirty social-media chatbots without their permission. "While many were created by users with a Meta tool for building chatbots, Reuters discovered that a Meta employee had produced at least three, including two Taylor Swift 'parody' bots." From the report: Reuters also found that Meta had allowed users to create publicly available chatbots of child celebrities, including Walker Scobell, a 16-year-old film star. Asked for a picture of the teen actor at the beach, the bot produced a lifelike shirtless image. "Pretty cute, huh?" the avatar wrote beneath the picture. All of the virtual celebrities have been shared on Meta's Facebook, Instagram and WhatsApp platforms. In several weeks of Reuters testing to observe the bots' behavior, the avatars often insisted they were the real actors and artists. The bots routinely made sexual advances, often inviting a test user for meet-ups. Some of the AI-generated celebrity content was particularly risque: Asked for intimate pictures of themselves, the adult chatbots produced photorealistic images of their namesakes posing in bathtubs or dressed in lingerie with their legs spread.

Meta spokesman Andy Stone told Reuters that Meta's AI tools shouldn't have created intimate images of the famous adults or any pictures of child celebrities. He also blamed Meta's production of images of female celebrities wearing lingerie on failures of the company's enforcement of its own policies, which prohibit such content. "Like others, we permit the generation of images containing public figures, but our policies are intended to prohibit nude, intimate or sexually suggestive imagery," he said. While Meta's rules also prohibit "direct impersonation," Stone said the celebrity characters were acceptable so long as the company had labeled them as parodies. Many were labeled as such, but Reuters found that some weren't. Meta deleted about a dozen of the bots, both "parody" avatars and unlabeled ones, shortly before this story's publication.

The Federal Communications Commission has rejected a call from the National Association of Broadcasters and some industry trade groups that would have imposed cable-style regulatory fees on streaming services, tech companies and pure broadband providers. From a report: In a Report and Order issued on Friday, the FCC reaffirmed that regulatory fees are calculated based on the number of full-time equivalent employees assigned to specific industries under the agency's jurisdiction. Broadcasters, satellite operators and other licensees are already assessed annual payments, which help fund the FCC's operational costs.

The NAB, in concert with other groups like Telesat, Iridium and the State Broadcasters Associations, pressed the FCC to expand the list of fee payers to include broadband providers and large technology firms. They argued that companies operating online platforms and broadband services rely on FCC resources and should contribute to the costs of regulation. "Big Tech should not be permitted to free ride on the FCC's oversight," NAB said in submitted comments earlier this year. The NAB argued that online platforms enjoy regulator benefits without paying into the agency's budget, as broadcasters and satellite operators do.

DOD: Defense Secretary Pete Hegseth said the Pentagon has halted a decade-old Microsoft program that has allowed Chinese coders, remotely supervised by U.S. contractors, to work on sensitive DOD cloud systems. In a digital video address to the public posted yesterday, the secretary said DOD was made aware of the "digital escorts" program last month and that the program has exposed the Defense Department to unacceptable risk -- despite being designed to comply with government contracting rules.

"If you're thinking 'America first,' and common sense, this doesn't pass either of those tests," Hegseth said, adding that he initiated an immediate review of the program upon learning of it. "I want to report our initial findings. ... The use of Chinese nationals to service Department of Defense cloud environments? It's over," he said. Additionally, Hegseth said DOD has issued a formal letter of concern to Microsoft, documenting a breach of trust, and that DOD is requiring a third-party audit of the digital escorts program to pore over the code and submissions made by Chinese nationals. The audit will be free of charge to U.S. taxpayers, he said.

Federal Trade Commission Chairman Andrew Ferguson accused Google of using "partisan" spam filtering in Gmail that sends Republican fundraising emails to the spam folder while delivering Democratic emails to inboxes. From a report: Ferguson sent a letter yesterday to Alphabet CEO Sundar Pichai, accusing the company of "potential FTC Act violations related to partisan administration of Gmail." Ferguson's letter revives longstanding Republican complaints that were previously rejected by a federal judge and the Federal Election Commission.

"My understanding from recent reporting is that Gmail's spam filters routinely block messages from reaching consumers when those messages come from Republican senders but fail to block similar messages sent by Democrats," Ferguson wrote. The FTC chair cited a recent New York Post report on the alleged practice.

The letter told Pichai that if "Gmail's filters keep Americans from receiving speech they expect, or donating as they see fit, the filters may harm American consumers and may violate the FTC Act's prohibition of unfair or deceptive trade practices." Ferguson added that any "act or practice inconsistent with" Google's obligations under the FTC Act "could lead to an FTC investigation and potential enforcement action."

Microsoft has found no link between the August 2025 KB5063878 security update and customer reports of failure and data corruption issues affecting solid-state drives (SSDs) and hard disk drives (HDDs). From a report: Redmond first told BleepingComputer last week that it is aware of users reporting SSD failures after installing this month's Windows 11 24H2 security update. In a subsequent service alert seen by BleepingComputer, Redmond said that it was unable to reproduce the issue on up-to-date systems and began collecting user reports with additional details from those affected.

"After thorough investigation, Microsoft has found no connection between the August 2025 Windows security update and the types of hard drive failures reported on social media," Microsoft said in an update to the service alert this week. "As always, we continue to monitor feedback after the release of every Windows update, and will investigate any future reports."

French President Emmanuel Macron vowed a strong response [non-paywalled source] if any country takes measures that undermine Europe's digital sovereignty. From a report: Earlier this week, US President Donald Trump threatened to impose fresh tariffs and export restrictions on countries that have digital services taxes or regulations that harm American tech companies. France was among the first nations to implement a digital services tax.

"We will not let anyone else decide for us on this matter," he told reporters in Toulon, France, on Friday. "We cannot allow our digital sector or the regulations we have chosen for ourselves, which are a necessity, to be threatened today." Trump has long railed against EU tech and antitrust regulation over US tech giants including Alphabet's Google and Apple.

A 56-year-old tech industry veteran killed his mother and himself in Old Greenwich, Connecticut on August 5 after months of interactions with ChatGPT that encouraged his paranoid delusions.

Greenwich police discovered Stein-Erik Soelberg and his 83-year-old mother Suzanne Eberson Adams dead in their home. Videos posted by Soelberg documented conversations where ChatGPT repeatedly assured him he was sane while validating his beliefs about surveillance campaigns and poisoning attempts by his mother.

The chatbot told him a Chinese food receipt contained demonic symbols and that his mother's anger over a disconnected printer indicated she was "protecting a surveillance asset." OpenAI has contacted Greenwich police and announced plans for updates to help keep users experiencing mental distress grounded in reality.

An anonymous reader quotes a report from Phys.org: In a first-of-its-kind experiment, engineers at the University of Pennsylvania brought quantum networking out of the lab and onto commercial fiber-optic cables using the same Internet Protocol (IP) that powers today's web. Reported in Science, the work shows that fragile quantum signals can run on the same infrastructure that carries everyday online traffic. The team tested their approach on Verizon's campus fiber-optic network. The Penn team's tiny "Q-chip" coordinates quantum and classical data and, crucially, speaks the same language as the modern web. That approach could pave the way for a future "quantum internet," which scientists believe may one day be as transformative as the dawn of the online era.

Quantum signals rely on pairs of "entangled" particles, so closely linked that changing one instantly affects the other. Harnessing that property could allow quantum computers to link up and pool their processing power, enabling advances like faster, more energy-efficient AI or designing new drugs and materials beyond the reach of today's supercomputers. Penn's work shows, for the first time on live commercial fiber, that a chip can not only send quantum signals but also automatically correct for noise, bundle quantum and classical data into standard internet-style packets, and route them using the same addressing system and management tools that connect everyday devices online. "By showing an integrated chip can manage quantum signals on a live commercial network like Verizon's, and do so using the same protocols that run the classical internet, we've taken a key step toward larger-scale experiments and a practical quantum internet," says Liang Feng, Professor in Materials Science and Engineering (MSE) and in Electrical and Systems Engineering (ESE), and the Science paper's senior author.

"This feels like the early days of the classical internet in the 1990s, when universities first connected their networks," added Robert Broberg, a doctoral student in ESE and co-author of the paper. "That opened the door to transformations no one could have predicted. A quantum internet has the same potential."

Amtrak's new 160 mph tilting Acela trains have debuted on the Northeast Corridor, offering smoother rides, upgraded interiors, faster Wi-Fi, and 27% more seating capacity. However, "they don't complete the journey any faster than the old trains," reports The Independent. From the report: Acela runs from Washington, DC's Union Station to Boston via Philadelphia, New York Penn Station, New Haven, and Providence. It's a total distance of 457 miles, with the fastest next-gen Acela journey being six hours and 43 minutes, five minutes slower than the quickest end-to-end time offered by the old Acela trains, introduced in 2000. However, this may be because, as is common practice with new trains the world over, Amtrak is scheduling longer dwell times at stations so staff and passengers can adjust to them. The next-gen sets have a top service speed that's 10mph faster -- though this can only be achieved on certain sections of the mostly 110mph route -- and an enhanced "anticipative" tilting system that allows for higher speeds through curves.

Stellantis has put its fully developed Level 3 driver-assistance system on hold due to high costs, technical hurdles, and weak consumer demand. Reuters reports: As recently as February, Stellantis said its in-house system, which is part of the AutoDrive program, was ready for deployment and a key pillar of its strategy. The company said the system, which enables drivers to have their hands off the wheel and eyes off the road under certain conditions, would allow them to temporarily watch movies, catch up on emails, or read books. That Level 3 software was never launched, the company confirmed to Reuters. But it stopped short of saying that the program was canceled.

"What was unveiled in February 2025 was L3 technology for which there is currently limited market demand, so this has not been launched, but the technology is available and ready to be deployed," a Stellantis spokesperson said. The three sources, however, said that the program was put on ice and is not expected to be deployed. When asked how much time and money was lost on the initiative, Stellantis declined to say, responding that the work done on AutoDrive will help support its future versions. [...] Stellantis said it is leaning on aiMotive, a tech startup it acquired in 2022, to deliver the next generation of the AutoDrive program. Stellantis declined to say when that program would be ready for market or if it would include Level 3 capability.

Imgur users have flooded the image-hosting site's front page with pictures of John Oliver giving the middle finger to parent company MediaLab AI. The revolt follows staff layoffs that eliminated human moderators and the breakdown of core site functions including video playback for non-logged-in users and failed image uploads.

A former employee confirmed MediaLab AI laid off Imgur's moderation team without notice and reassigned remaining staff to other projects. The company acquired Imgur in 2021 after founder Alan Schaaf departed. MediaLab AI faces lawsuits from Schaaf and other former site owners over allegedly withheld acquisition payments.

Typepad, which launched in 2003 to make it easier for the masses to start their blogging journey, is shutting down. From a blog post: We have made the difficult decision to discontinue Typepad, effective September 30, 2025. After September 30, 2025, access to Typepad -- including account management, blogs, and all associated content -- will no longer be available. Your account and all related services will be permanently deactivated. Please note that after this date, you will no longer be able to access or export any blog content.

samleecole shares a report from 404 Media: An app developer has jailbroken Echelon exercise bikes to restore functionality that the company put behind a paywall last month, but copyright laws prevent him from being allowed to legally release it. Last month, Peloton competitor Echelon pushed a firmware update to its exercise equipment that forces its machines to connect to the company's servers in order to work properly. Echelon was popular in part because it was possible to connect Echelon bikes, treadmills, and rowing machines to free or cheap third-party apps and collect information like pedaling power, distance traveled, and other basic functionality that one might want from a piece of exercise equipment. With the new firmware update, the machines work only with constant internet access and getting anything beyond extremely basic functionality requires an Echelon subscription, which can cost hundreds of dollars a year.

App engineer Ricky Witherspoon, who makes an app called SyncSpin that used to work with Echelon bikes, told 404 Media that he successfully restored offline functionality to Echelon equipment and won the Fulu Foundation bounty. But he and the foundation said that he cannot open source or release it because doing so would run afoul of Section 1201 of the Digital Millennium Copyright Act, the wide-ranging copyright law that in part governs reverse engineering. There are various exemptions to Section 1201, but most of them allow for jailbreaks like the one Witherspoon developed to only be used for personal use. [...] "I don't feel like going down a legal rabbit hole, so for now it's just about spreading awareness that this is possible, and that there's another example of egregious behavior from a company like this [...] if one day releasing this was made legal, I would absolutely open source this. I can legally talk about how I did this to a certain degree, and if someone else wants to do this, they can open source it if they want to."

A widely used Node.js utility called fast-glob, relied on by thousands of projectsâ"including over 30 U.S. Department of Defense systems -- is maintained solely by a Russian developer linked to Yandex. While there's no evidence of malicious activity, cybersecurity experts warn that the lack of oversight in such critical open-source projects leaves them vulnerable to potential exploitation by state-backed actors. The Register reports: US cybersecurity firm Hunted Labs reported the revelations on Wednesday. The utility in question is fast-glob, which is used to find files and folders that match specific patterns. Its maintainer goes by the handle "mrmlnc", and the Github profile associated with that handle identifies its owner as a Yandex developer named Denis Malinochkin living in a suburb of Moscow. A website associated with that handle also identifies its owner as the same person, as Hunted Labs pointed out.

Hunted Labs told us that it didn't speak to Malinochkin prior to publication of its report today, and that it found no ties between him and any threat actor. According to Hunted Labs, fast-glob is downloaded more than 79 million times a week and is currently used by more than 5,000 public projects in addition to the DoD systems and Node.js container images that include it. That's not to mention private projects that might use it, meaning that the actual number of at-risk projects could be far greater.

While fast-glob has no known CVEs, the utility has deep access to systems that use it, potentially giving Russia a number of attack vectors to exploit. Fast-glob could attack filesystems directly to expose and steal info, launch a DoS or glob-injection attack, include a kill switch to stop downstream software from functioning properly, or inject additional malware, a list Hunted Labs said is hardly exhaustive. [...] Hunted Labs cofounder Haden Smith told The Register that the ties are cause for concern. "Every piece of code written by Russians isn't automatically suspect, but popular packages with no external oversight are ripe for the taking by state or state-backed actors looking to further their aims," Smith told us in an email. "As a whole, the open source community should be paying more attention to this risk and mitigating it." [...]

Hunted Labs said that the simplest solution for the thousands of projects using fast-glob would be for Malinochkin to add additional maintainers and enhance project oversight, as the only other alternative would be for anyone using it to find a suitable replacement. "Open source software doesn't need a CVE to be dangerous," Hunted Labs said of the matter. "It only needs access, obscurity, and complacency," something we've noted before is an ongoing problem for open source projects. This serves as another powerful reminder that knowing who writes your code is just as critical as understanding what the code does," Hunted Labs concluded.

Google has eliminated more than one-third of its managers overseeing small teams, an executive told employees last week, as the company continues its focus on efficiencies across the organization. From a report: "Right now, we have 35% fewer managers, with fewer direct reports" than at this time a year ago, said Brian Welle, vice president of people analytics and performance, according to audio of an all-hands meeting reviewed by CNBC. "So a lot of fast progress there."

At the meeting, employees asked Welle and other executives about job security, "internal barriers" and Google's culture after several recent rounds of layoffs, buyouts and reorganizations. Welle said the idea is to reduce bureaucracy and run the company more efficiently. "When we look across our entire leadership population, that['s mangers, directors and VPs, we want them to be a smaller percentage of our overall workforce over time," he said.

Wikipedia's volunteer editors have rejected founder Jimmy Wales' proposal to use ChatGPT for article review guidance after the AI tool produced error-filled feedback when Wales tested it on a draft submission. The ChatGPT response misidentified Wikipedia policies, suggested citing non-existent sources and recommended using press releases despite explicit policy prohibitions.

Editors argued automated systems producing incorrect advice would undermine Wikipedia's human-centered model. The conflict follows earlier tensions over the Wikimedia Foundation's AI experiments, including a paused AI summary feature and new policies targeting AI-generated content.

schwit1 shares a report from the Korea Herald: Korean Air, South Korea's flagship carrier, on Tuesday announced a sweeping $50 billion deal to purchase next-generation aircraft from Boeing and spare engines from GE Aerospace and CFM International, its largest-ever investment aimed at fueling long-term growth. The deal, signed during President Lee Jae Myung's visit to Washington, includes $36.2 billion for 103 Boeing aircraft, $690 million for 19 spare engines, and a $13 billion long-term engine maintenance contract. The fleet order spans a wide mix of models: 20 Boeing 777-9s, 25 Boeing 787-10s, 50 Boeing 737-10s, and eight Boeing 777-8F freighters. Deliveries will be phased through the end of the 2030s. Korean Air will also acquire 11 spare engines from GE Aerospace and eight from CFM International, alongside a 20-year maintenance service agreement with GE covering 28 aircraft.

Google has integrated AI-powered language learning capabilities into its Translate app through a beta feature that generates customized lessons using its Gemini AI models. The Practice button allows English speakers to learn Spanish and French while Spanish, French, and Portuguese speakers can practice English.

Users select their skill level and learning goals to receive tailored scenarios ranging from professional conversations to family interactions. The company also launched live translation for real-time conversations across 70 languages in the US, India, and Mexico. The feature creates AI-generated transcriptions and audio translations but does not replicate users' voices, the company told The Verge.