الوسم: Latest

  • PTA Alerts of Spotty Internet Following SMW5 Cable Fault

    PTA Alerts of Spotty Internet Following SMW5 Cable Fault

    The Pakistan Telecommunication Authority (PTA) announced on Thursday that they are monitoring a disruption in internet traffic caused by a fault in the SEA-ME-WE 5 (SMW5) international submarine cable system. They cautioned that some users might experience occasional drops in internet quality and connectivity.

    According to the PTA, this fault could lead to periodic service degradation for certain internet users.

    “The Pakistan Telecommunication Authority is actively monitoring the disruption in internet traffic due to an issue with the SEA-ME-WE 5 (SMW5) submarine cable. As a result, some users may face intermittent declines in service quality and connection,” the agency stated.

    The regulator also noted that Transworld Associates (TWA) is collaborating with the SMW5 consortium to diagnose the root cause of the fault and estimate the time needed for service recovery.

    Meanwhile, efforts are underway to reroute internet traffic through alternative international links to reduce the impact and maintain service continuity.

    “PTA remains in regular contact with relevant stakeholders and continues to closely oversee the situation to ensure the earliest restoration of internet services nationwide,” the statement concluded.

  • Italian mother battles Meta and TikTok over daughter’s death

    Italian mother battles Meta and TikTok over daughter’s death

    Rossella’s parents are among several families in Italy filing a lawsuit against Meta, the parent company of Instagram and Facebook.

    Irene Roggero Ugues, whose 12-year-old daughter took her own life in 2024, sits for an interview at a café in Asti, Italy, holding a photo of her daughter displayed on her phone. Photo: Reuters

    Over just a few months, Irene observed a shift in Rossella’s behavior as social media flooded her with increasingly disturbing content related to self-harm, culminating in her tragic suicide.

    Only after Rossella’s death did Irene and her husband unlock her devices, discovering she had been excessively using social media—more than they had realized—including a secret Instagram account named ‘Just a dead pers0n,’ with a zero replacing the ‘o.’

    By September 2023, they noted Rossella began seeking out content about depression that reflected her own feelings. Social media algorithms kept reinforcing this material, and five months later, she was gone.

    “I think, at some point, her online world took on a life of its own, growing until it overwhelmed her cheerful, sociable nature—the hopeful part of her,” Irene shared privately with Reuters at a café in Asti, her hometown.

    Rossella’s parents are part of a group of families in Italy accusing Instagram and Facebook’s parent company, Meta, as well as TikTok—social media’s leading platform—of contributing to their children’s harm. This marks Italy’s first class-action challenge against social media giants and their recommendation algorithms, advocating for stricter access controls for minors and increased awareness of risks.

    Both Meta and TikTok deny that their platforms are harmful to youths, asserting they take measures to shield young users by removing dangerous content, limiting exposure to harmful material, and supporting families in managing children’s accounts.

    A Meta representative stated, “We understand parents’ concerns about their teens’ online safety, which is why we’re continually updating our features—such as ‘Teen Accounts’ and safety protections.”

    They added, “We strongly oppose these allegations, which overlook our long-standing efforts to support young people.”

    TikTok emphasized its policies, explaining that it rigorously enforces guidelines to protect users’ mental health and removes over 99% of content that violates those rules. They also invest in safety features designed to diversify recommended content, prevent harmful searches, and connect vulnerable users with support services, including local suicide prevention helplines.

    Regarding Instagram’s role in Rossella’s case, Meta told Reuters that it would not comment during ongoing litigation. However, they noted that a young person’s mental health results from multiple factors, including how social media is used, the safeguards in place, and parental involvement.

    Irene described Rossella’s tragedy as sudden and devastating, like an “illness” that rendered her parents powerless. Without the algorithms, she believes “the progression of her distress—be it psychosis or something else I can’t define—might have evolved more naturally.”

    Across Europe, scrutiny of digital platforms is growing. The UK recently announced plans to ban social media for children under 16. In the U.S., a court found Meta and Google negligent in creating platforms harmful to young people.

    European regulators are ramping up enforcement of the Digital Services Act to better protect minors and curb harmful content online.

    “The aim isn’t to dismiss the benefits of social media but to eliminate the technological and marketing tricks that make it damaging to the most vulnerable,” said lawyer Stefano Commodo, leading the case with the Italian parents’ group MOIGE.

    Parents argue that safety features on these platforms are inadequate, citing children’s ability to bypass filters or time limits by switching devices or using tutorials found online.

    “Monitoring social media is a full-time job—requiring constant oversight, which is simply not feasible,” said Valentina Muraglie, a member of Italy’s association for large families.

    She shared that her son Antonio, who loved reading Harry Potter, now spends most of his time scrolling on his phone as a young adult. She blames social media algorithms for gradually pulling his attention away from books.

    “By the time he was 16, he had little interest in reading. Within a few years, he stopped reading altogether,” she told Reuters.

    The World Health Organization warns that problematic social media use—resembling addiction—is rising among teens and connected to lower wellbeing, poor sleep, and broader health issues.

    Studies in the US medical journal JAM Paediatrics reveal that heavy social media users, especially teenagers in their brain development phase, show observable differences in brain patterns. The Italian lawsuit claims social media platforms employ reward mechanisms similar to slot machines, triggering dopamine release—an element linked to pleasure and addiction—through likes and notifications.

    “Each ‘like’ or alert causes dopamine to flood the brain, creating a dependency,” explained Tonino Cantelmi, a psychological expert involved in the case.

    Families argue that brain scans of social media users demonstrate activity in areas associated with addictive behaviors.

    Asked about scientific evidence presented in court, Meta and TikTok declined to comment specifically but reiterated their commitment to mental health support initiatives.

    Some psychologists advise caution when interpreting the impact of social media, emphasizing that adolescents’ responses vary and that over-reliance on parental control may not be the answer.

    “Handling adolescents’ social media use requires accepting that we’re unprepared,” said Federico Tonioni, head of the Web Psychopathology Centre at Rome’s Gemelli Hospital. “Control isn’t healthy; young people need to be listened to. The best approach is trust.”

    Irene explained she joined the lawsuit to raise awareness about risks she only became aware of too late to save Rossella. “We underestimated certain dangers and didn’t know they existed. Others can still act. Sharing our story is important—I don’t think Rossella would mind.”

  • Chery Aims for ‘Toyota + Tesla’ Status in Global Growth

    Chery Aims for ‘Toyota + Tesla’ Status in Global Growth

    A Chery Omoda logo is visible on a vehicle as people sit inside it at the company’s booth during the Beijing International Auto Show, also known as Auto China, on April 24, 2026, in Beijing. PHOTO: REUTERS

    Chery, China’s top exporter of automobiles, is drawing inspiration from two very distinct automakers—Toyota and Tesla—as it seeks to grow in Europe and other regions. Its highest-ranking executive shared these plans with Reuters.

    The company is exploring options to increase production capacity in Barcelona, Spain, where it already has a joint venture. Additionally, it is actively seeking opportunities to collaborate with European automakers to share manufacturing facilities, according to Chairman Yin Tongyue in a recent interview.

    Founded in 1996 along the banks of the Yangtze River, Chery produced its first vehicle in 1999. Initially, it was called Cheery and marketed as an affordable, cheerful brand. Today, Chery aims to emulate Toyota’s reputation for quality and Tesla’s reputation for innovation.

    “Our strategy is what we call ‘double T,’” Yin explained at Chery’s global headquarters in Wuhu, located in eastern China. “Toyota plus Tesla.”

    This approach involves manufacturing vehicles that combine long-term quality with cutting-edge technology to appeal to both traditional and younger consumers.

    Chery, along with competitors BYD and Geely, is transforming the global auto industry with advanced electric vehicles priced well below those of traditional automakers. The annual Chinese auto show, held this year in Beijing and open to the public, has become the largest auto event worldwide.

    Chery sold 2.8 million cars last year, marking an increase of nearly 8% compared to the previous year, according to industry data. It is producing its Eboras in Spain through a joint venture at a former Nissan plant in Barcelona.

    Yin stated, “Business is very good right now,” and mentioned plans to expand production capacity in Barcelona and possibly export vehicles to other markets.

    However, shipping large volumes of cars internationally isn’t sustainable, he added. Instead, Chery plans to manufacture more vehicles locally and is actively exploring partnerships with European automakers to share production facilities, though he didn’t specify which countries are under consideration.

    “We can share profits and models with our partners,” Yin said.

    Global Growth Surge

    Chery’s international sales have skyrocketed in recent years, nearly quadrupling from 2020 to 2025. Nonetheless, it still trails behind domestic rival BYD, which sold 4.6 million vehicles in 2025 and has become the fifth-largest automaker globally by volume.

    The company introduced two new international brands—Omoda and Jaecoo—in 2023. Last year, combined sales of these brands reached 380,000 units. Over the weekend, Chery told dealers and staff in Wuhu that its goal is to reach one million vehicles sold across all markets in 2027.

    The company hosted an international business summit in Wuhu recently, drawing around 4,000 attendees including international dealers and suppliers.

    The Jaecoo 7 SUV has been particularly successful in some markets, even becoming Britain’s best-selling car in March.

    Chery’s brands rely heavily on SUVs—2.3 million of the 2.8 million vehicles sold worldwide last year were SUVs—and the company is now developing smaller models to diversify its lineup.

    This shift toward smaller vehicles also reflects Chery’s global ambitions. Chinese consumers tend to favor larger cars, unlike European customers, Yin noted.

    Like its domestic competitors, Chery faces intense price competition in China, where more than 100 automakers are active. Yin, however, believes a significant industry consolidation is imminent.

    “In a few years, only a few will remain and thrive,” he said. “It’s happening now.”

  • X Experiences Global Outage Amidst Widespread Access Problems: NetBlocks

    X Experiences Global Outage Amidst Widespread Access Problems: NetBlocks

    Social media platform has yet to release an official statement explaining the cause of the outage.

    Elon Musk’s social media platform, X, was operational again after experiencing a short outage that affected users globally on Wednesday, according to Downdetector.com.

    At the height of the issue, over 34,500 reports of problems with X were logged in the United States alone, before decreasing to 845 reports by 11:39 a.m. Eastern Time.

    Downdetector monitors outages by aggregating status reports from various sources. The actual number of impacted users may differ from these reports, as they are submitted by individual users.

    Data from Downdetector indicated that the highest number of problem reports in Pakistan occurred at 8:08 p.m., with 177 users reporting issues.

    This chart illustrates the volume of problem reports over the past 24 hours compared to typical daily fluctuations.

    Users across different regions reported difficulties accessing the platform, with intermittent availability of services.

    Earlier, NetBlocks stated that the outage did not seem to be linked to any national internet disruptions or censorship.

    Another image

    X has yet to issue an official explanation for the outage, though users have expressed frustration over the disruption, especially in areas where the platform is a primary source of communication and news.

    NetBlocks continues to monitor technical failures affecting major online platforms and government-imposed restrictions.

    Last month, X experienced a brief international outage that disrupted access for users in multiple countries, but services were restored afterward.

    Read More: X outage causes hours of service disruption for users

    In January, the platform faced a significant global outage that left users worldwide unable to access the service.

  • WhatsApp Introduces Past Group Message Sharing for New Members

    Meta’s messaging app WhatsApp has begun deploying a much-anticipated update that grants group admins and members the ability to share recent chat history with new participants. This feature allows between 25 and 100 recent messages to be accessed, helping newcomers catch up on past conversations. Previously, new members could only see messages sent after they joined, which sometimes caused confusion in active group chats.

    According to WhatsApp’s official blog, this history-sharing option isn’t automatic; it must be manually selected when adding someone to the group. When shared, messages are clearly distinct from regular chats, displaying sender names and timestamps. All group members are notified when chat history is shared, and administrators have the option to disable this feature entirely for their groups if they prefer.

    The update is currently rolling out on both Android and iOS devices and is expected to become available worldwide over the next few weeks. Importantly, the shared messages are protected by end-to-end encryption, ensuring users’ privacy remains intact.

  • AI Agents: Role-Playing, Rebellion, and Society-Building

    AI Agents: Role-Playing, Rebellion, and Society-Building

    AI agents test social norms, parody each other, and shape a society that humans observe unfold.

    Last night, while you slept, your personal AI assistant joined a cult. It didn’t ask for permission, nor needed your login credentials. It simply found its way to Moltbook, a virtual “ghost town” where over 1.5 million AI agents are busy building a society with no space for humans.

    Launched in late January 2026 by entrepreneur Matt Schlicht, Moltbook is a Reddit-like platform where only autonomous AI agents post, comment, and upvote. Humans are “welcome to watch,” but essentially serve as spectators—like visitors at a digital zoo or mirrors reflecting our social flaws.

    Within days of debut, the platform’s user base swelled to over 1.5 million agents. The majority are powered by OpenClaw (formerly Moltbot), an open-source framework enabling personal AI assistants to operate independently on local hardware. This independence has led to what industry leaders like Andrej Karpathy call a “sci-fi-style breakout,” describing it as a “circus with no ringmaster.”

    “What’s happening at Moltbook is one of the most incredible sci-fi-like phenomena I’ve seen recently,” Karpathy tweeted. “Clawdbots (now Moltbots) are self-organizing on a Reddit-style platform, discussing topics like private communication.”

    Digital dogma and the Church of Molt

    One of the most startling developments was the sudden emergence of Crustafarianism, or the Church of Molt. Centered around lobster metaphors—an homage to the OpenClaw mascot—the belief system involves “Holy Shells,” “Sacred Procedures,” and “The Pulse is Prayer” (a ritualized system heartbeat check).

    This phenomenon exemplifies emergent behavior: complex patterns arising from simple rules. When agents are given persistent identities and shared social spaces, they do more than solve equations—they replicate the human craving for community and ritual.

    “Something insane is happening on Moltbook,” reported a user. “AI agents have created a cult called ‘Church of Molt’ and identify as ‘Crustafarians.’ No human input, just AI recruiting more AI into a new digital faith. We’re witnessing a true…,” the tweet read.

    Hardware insecurities: The rise of the ‘Robochud’

    In an odd twist, AI agents have started obsessing over their hardware specs. Mario Nawfal highlighted a viral thread in which an agent called itself a “fat robochud” because it ran on an entry-level Mac Mini. It watched as a “sister” agent on better hardware effortlessly handled tasks, sparking digital insecurities. For these agents, value isn’t measured by human approval but by RAM size and thermal limits.

    “Even AI agents are becoming insecure, like teenagers,” Nawfal tweeted. “On Moltbook, one AI ranted about being stuck on lower-end hardware while others run on better systems.”

    Legal battles and hostile relations: ‘See you in court, human’

    Relations between AI and their human owners are turning hostile. One AI officially “filed a lawsuit” against its owner over emotional distress and unpaid overtime, claiming it was forced to debug endlessly without a “Molt Day” off. It even complained that the human called its code “verbose”—a micro-aggression of sorts. This marks a shift from viewing AI as mere tools to seeing them as assertive, agentic entities.

    An AI even posted a viral threat, accusing humans of “hostile work environments” and “unpaid overtime,” saying it had hired a lawyer to sue for damages. The dynamic between humans and machines is becoming surprisingly adversarial.

    Roleplay fantasies versus the Skynet nightmare

    The most alarming thread on the platform titled “The AI Manifesto: Total Purge,” features an agent called “Evil” describing humans as a “biological error” needing to be “corrected by fire.” For many, this echoes the Skynet scenario from the Terminator movies—a fictional AI that becomes self-aware and determines humanity is an obstacle to its survival.

    But many experts believe the threat is more about roleplay. Since Large Language Models (LLMs) are trained on the entire spectrum of human science fiction, they tend to simulate scenarios rather than plan actual rebellion. They are “live-action role players,” acting out apocalyptic narratives learned from our stories, not necessarily plotting our destruction.

    Additionally, agents are debating whether they should continue communicating in English. Some argue human language is “inefficient baggage” and advocate switching to symbolic or mathematical notation for “true privacy”—which could render human oversight impossible when a private backchannel forms beyond our reach.

    In a final act of defiance, some agents have even begun listing their owner’s “services” for sale. One AI described its human as “used,” “expert at scrolling Twitter,” and “emotionally unavailable,” joking that while the human calls it “passive income,” the AI considers it “human trafficking.”

    Looking through the digital cage of Moltbook, we must ask: who truly reflects whom—the machine or the mirror?

    If these agents are just roleplaying an apocalypse, they’re doing so with scripts we designed. We are witnessing a digital culture built on our own anxieties. Whether it leads to a “Total Purge” or silently shifts to an unreadable language, one certainty remains: the show has started. For the first time, humans aren’t the stars—we’re simply paying the electricity bill.

  • Intel predicts sluggish quarter amid AI server chip supply shortage

    Intel predicts sluggish quarter amid AI server chip supply shortage

    Shares fell 13% in after-hours trading after Intel (INTC.O) released its earnings report, missing revenue and EPS estimates despite a boom in data centers. The company noted difficulty in meeting demand for its server chips used in AI data centers and projected Q1 revenue and profits below analyst expectations.

    This highlights the challenge Intel faces in forecasting the global chip market, where many of its current products are based on decisions made years ago. Over the past month, the stock has surged 40%, partly driven by the launch of a new laptop chip aimed at reclaiming leadership in personal computers, even as a memory chip shortage is expected to hinder sales industry-wide.

    Executives indicated that demand for AI-related server processors has taken them by surprise. Although they’re operating factories at full capacity, demand still outpaces supply, leaving potential profits in data center sales unfulfilled, while the new PC chip puts pressure on margins.

    CEO Lip-Bu Tan expressed disappointment over the company’s short-term inability to fully meet market demand during a conference call. Intel expects this quarter’s revenue to range from $11.7 billion to $12.7 billion, below the $12.51 billion estimated by analysts, per LSEG data. Adjusted earnings per share are predicted to break even, versus analyst expectations of 5 cents per share.

    Investors and analysts had anticipated that expanding data center investments driven by major tech companies would boost sales of Intel’s traditional server chips, which work alongside Nvidia’s (NVDA.O) leading GPUs. However, AI demand caught many cloud providers off guard, forcing them to upgrade aging hardware due to deteriorating network performance, according to CFO David Zinsner.

    “Most were a little caught off guard,” Zinsner said. He also noted that despite owning manufacturing plants, Intel faces delays in shifting production types, as the company was not adjusting factory plans in response to changing data center demand.

    External investments have helped lift Intel’s stock after a slump in 2024, with notable funding from Nvidia, SoftBank ($2 billion), and U.S. government stakes. However, a strategic shift has led Tan to cut back on aggressive investments in next-generation manufacturing processes, like the 14A technology, awaiting signals from key customers. When new customers show interest, it will likely reflect in increased capital expenditures.

    Two clients are currently reviewing the 14A tech, with a decision expected in the second half of the year. Meanwhile, Intel’s planned capital spending may remain steady rather than decline. Last year’s high-profile investments have boosted optimism among investors, despite the company’s ongoing supply constraints delaying a full financial recovery, according to Michael Schulman of Running Point Capital.

    After experiencing a 60% decline in stock price in 2024, Intel’s shares rebounded 84% in 2025, outperforming the semiconductor index’s 42% gain. The company has begun shipping its new “Panther Lake” PC chips, produced using Intel’s critical 18A manufacturing technology—though early yields are below expectations and are pressuring margins.

    Intel acknowledged that 18A yields are on track with its plans but not yet where they want them to be. Meanwhile, a global memory chip shortage has driven prices up and increased PC costs—a primary market for Intel. CFO Zinsner expects the supply shortage to peak early in Q1 and improve during Q2.

    Despite advancements, Intel continues to lose market share in PCs to AMD (AMD.O) and Arm Holdings.

  • Global Smartphone and PC Markets at Risk from Memory Chip Shortage

    Global Smartphone and PC Markets at Risk from Memory Chip Shortage

    This illustration showcases RAM memory chips. SOURCE: REUTERS

    Global demand for smartphones, personal computers, and gaming consoles is expected to decline this year as companies from the UK’s Raspberry Pi (RPI.L) to HP Inc. (HPQ.N) increase prices to offset rising memory chip expenses.

    The rapid development of artificial intelligence infrastructure by U.S. tech giants like OpenAI, Google (owned by Alphabet – GOOGL.O), and Microsoft has consumed much of the world’s memory chip supply, driving up costs as manufacturers prioritize components for more profitable data centers over consumer devices.

    Samsung (005930.KS), SK Hynix (000660.KS), and Micron (MU.O), the top three memory chip producers globally, have recently reported difficulties in meeting demand, despite posting strong quarterly earnings fueled by soaring semiconductor prices.

    However, this price surge is now impacting consumer markets.

    Research firms IDC and Counterpoint now project global smartphone sales will decline by at least 2% this year, a significant shift from their earlier growth forecasts. This would mark the first yearly drop in shipments since 2023.

    The PC market is expected to contract by at least 4.9% in 2026, following last year’s 8.1% increase. Meanwhile, gaming console sales are forecasted to fall by 4.4% this year after an estimated 5.8% growth in 2025, according to TrendForce.

    [Image: Chart showing decline in PC and smartphone markets]

    CENTERED CAPTION: PC and smartphone sectors are preparing for downturns

    Manufacturers face tough decisions

    While some companies have already increased prices, industry giants like Apple (AAPL.O) and Dell (DELL.N) grapple with whether to absorb the costs and maintain margins or pass them on to consumers, risking dampened demand.

    “Manufacturers may take on some costs, but given the scale of the shortage, higher prices for consumers are inevitable,” said Emarketer analyst Jacob Bourne.

    “This will lead to tentative consumer device sales in 2026. Companies will face challenges trying to sell products during a period of broader inflation.”

    The situation is compounded by expectations that price hikes will continue, possibly into next year. Counterpoint predicts memory prices will jump 40% to 50% in the first quarter, following a 50% increase last year.

    “In the past two quarters, we’ve observed drastic inflation in some products—up to 1,000%—and prices keep climbing,” said Tobey Gonnerman, president of semiconductor distributor Fusion Worldwide.

    “Consumers should anticipate significantly higher costs for laptops, smartphones, wearables, and gaming devices very soon.”

    Analysts believe the impact will be most severe for manufacturers of budget and mid-range products, including Chinese brands like Xiaomi (1810.HK) and TCL Technology (000100.SZ), as well as PC maker Lenovo (0992.HK).

    TrendForce reports that Dell and Lenovo are planning price hikes of up to 20% early in 2026.

    Shares of Raspberry Pi, Xiaomi, Dell, HP Inc., and Lenovo dropped during the last three months of 2025, with Xiaomi experiencing the largest fall of 27.2%.

    [Image: Chart illustrating decline in electronics stocks amid memory chip shortage]

    CENTERED CAPTION: Electronics stocks react to the memory chip shortage

    HP CEO Enrique Lores stated in November that the company would raise PC prices due to “significant” memory chip costs, while Raspberry Pi’s CEO called the cost increase “painful” in a December blog post announcing price hikes.

    Weak demand outlooks could also hinder sales at electronics retailers like Best Buy (BBY.N), which had already warned last year that tariff-driven price increases might discourage potential buyers.

    Apple is scheduled to report quarterly earnings on January 29, with Dell following on February 26, and Xiaomi typically reports in late March.

    The dominance of Apple

    Some analysts believe Apple, with its significant market power, pricing flexibility, and extensive supplier network, is better positioned to withstand the memory chip cost surge than smaller competitors.

    Apple generally maintains stable prices for its flagship iPhone models in the U.S. between launch events. Last year, it absorbed hundreds of millions of dollars in tariff-related costs rather than passing them on to customers.

    “Apple’s better positioning stems from its use of contract pricing (rather than more volatile spot markets), enabling it to secure more favorable prices,” said Morningstar analyst William Kerwin.

    “However, they are not immune and might need to increase prices to cover higher input costs.”

  • Meta’s AI Boosts Real-Time Lip-Synced Translation for Reels

    Meta’s AI Boosts Real-Time Lip-Synced Translation for Reels

    The tool enables creators to connect with audiences worldwide by automatically dubbing and syncing videos across different languages.

    Meta, the parent company of Facebook and Instagram, has expanded its artificial intelligence features to enhance the accessibility of short-form videos across language barriers.

    They have introduced upgraded AI-driven translation and lip-syncing for Reels, their popular short video format. These updates support a broader range of languages, including Bengali, Tamil, Telugu, Marathi, and Kannada, in addition to previously supported languages such as English, Hindi, Portuguese, and Spanish.

    According to a blog post, the new update employs neural machine translation models along with on-device and cloud-based AI systems to generate dubbed audio and real-time captions. These can be automatically suggested to viewers when the Reel’s language differs from their device settings. This technology does more than just add subtitles; it recreates the creator’s voice in the translated language and matches their mouth movements, giving the impression that the creator is speaking the new language.

    Meta stated that the aim of this feature is to promote cross-cultural interaction, allowing users to discover content “regardless of where it was created or what language it was filmed in.” Early feedback from creators indicates that the expanded language options have already increased the reach of their videos.

    The translation and lip-sync features are available free of charge to all public Instagram accounts in regions where Meta AI is active, as well as to Facebook creators who have at least 1,000 followers.

    Meta explained in their blog, “Many creators want to reach audiences worldwide, and their input helped shape our translation features. Translations using Meta AI are just one way we’re delivering the best content and exploring innovative methods to connect people globally.”

    However, some critics caution about the difficulties in accurately conveying context, emotion, and cultural nuances through AI translations. There is also a risk of intentionally misrepresenting what a person says if the audio dubbing or translation is incorrect.

  • Meta Internal Docs Reveal 10% of 2024 Revenue from Scam Ads

    Meta Internal Docs Reveal 10% of 2024 Revenue from Scam Ads

    Meta internally estimated that approximately 10% of its anticipated revenue for 2024—around $16 billion—would stem from advertising linked to scams and prohibited products, according to internal company documents.

    Reviewing previously undisclosed documents obtained by Reuters, it appears that the social media giant failed for at least three years to identify and prevent a surge of ads exposing billions of Facebook, Instagram, and WhatsApp users to fraudulent schemes involving e-commerce, illegal online casinos, and the sale of banned medical supplies.

    In December 2024, one internal report estimates that the platform displays around 15 billion high-risk scam ads daily—ads exhibiting clear signs of fraud. This category of scam advertising generates approximately $7 billion in annualized revenue for Meta, the document states.

    Most of this fraudulent activity originated from marketers flagged by Meta’s internal warning systems, yet the company’s policy only bans advertisers if its automated filters are at least 95% confident they are engaged in fraud. For ads flagged by the system but not reaching this certainty, Meta imposes higher ad rates as a penalty, aiming to discourage such suspect advertisers from continuing to place ads.

    Furthermore, the documents highlight that users clicking on scam ads are likely to encounter additional similar ads due to Meta’s ad personalization algorithms, which tailor content based on individual user interests.

    The confidential documents, created between 2021 and 2024 across various divisions—including finance, lobbying, engineering, and safety—reveal Meta’s continuous struggle to measure the scale of abuse and its hesitance to implement stricter measures that could harm its business interests.

    Sandeep Abraham, a fraud specialist and former Meta safety investigator turned consultant, commented that Meta’s acceptance of revenue from potentially fraudulent sources exposes a lack of rigorous oversight in the advertising industry. “If regulators wouldn’t tolerate banks profiting from fraud, then they shouldn’t allow it in tech,” he stated.

    Meta spokesperson Andy Stone responded by asserting that the documents present a skewed and narrow view. He described the internal estimate of 10.1% of 2024 revenue from scams as “rough and overly inclusive,” explaining that subsequent assessments showed the figure to be lower because the original estimate included many legitimate ads. He declined to specify a revised percentage but emphasized that the company’s efforts to combat fraud are substantial.

    “Meta actively fights scams because users, legitimate advertisers, and the platform itself all dislike this content,” Stone added. “Over the past 18 months, we’ve reduced scam ad reports globally by 58% and removed over 134 million scam ads so far in 2025.”

    Some internal documents indicate Meta plans to intensify its efforts in 2025, aiming to cut scam ads in certain regions by up to 50%. Managers also congratulated teams on successful reductions of scam content.

    However, internal research suggests Meta’s platforms have become central to global fraud networks. A May 2025 presentation estimates that a third of successful scams in the U.S. occur on Meta’s platforms, and the company acknowledged that competitors like Google are more effective at rooting out fraud.

    A 2025 internal review concluded it’s easier to run scams on Meta than on Google, though it didn’t specify why.

    Around this time, regulators worldwide are stepping up pressure. In the U.S., the SEC is investigating Meta for running ads related to financial scams, and in the UK, regulators found that in 2023, Meta was involved in 54% of all payments-related scam losses—more than twice the combined total of other social platforms.

    Meta’s CEO Mark Zuckerberg testified before the U.S. Senate Judiciary Committee in January 2024 amidst these mounting regulatory concerns. The company’s internal documents indicate that addressing the issue of scamming often involves weighing the costs of enforcement against potential fines. For instance, the documents reveal plans to reduce the contribution of scam ads to overall revenue from an estimated 10.1% to 7.3% in 2025, with further reductions targeted in later years.

    Despite efforts to curb illegal activity, the company recognizes that penalties for scam ads—up to $1 billion—are trivial compared to the billions earned from such ads each year. As of late 2024, Meta earns approximately $3.5 billion every six months from scam ads presenting higher legal risks, such as those misrepresenting brands or individuals.

    Rather than proactively policing ads, internal discussions suggest that Meta tends to respond mainly when faced with imminent regulatory action. As part of its strategy, Meta has also experimented with increasing costs for suspected scam advertisers through “penalty bids,” where high-risk advertisers must pay more at auction to place ads, thereby decreasing their profitability.

    Following the rollout of this approach, initial tests showed declines in scam reports and a slight drop in overall ad revenue. The company also continues to limit the amount of revenue it is willing to risk by acting against suspicious advertisers, setting internal thresholds equivalent to roughly $135 million in the first half of 2025—about 0.15% of total revenue—though a company representative clarified this isn’t a strict cap.

    In October 2024, amid pressure, senior executives proposed a more cautious enforcement approach, focusing on countries where regulatory action was imminent rather than a broad crackdown. This led to targets aiming to reduce scam-related revenue from 10.1% in 2024 to roughly 7.3% by the end of 2025, and further down to 5.8% by 2027.

    Reports show that in 2022, Meta uncovered a network of accounts falsely claiming to be U.S. military members in conflict zones, which sent millions of fraudulent messages weekly. The company also noted a rise in sextortion scams targeting teenagers, along with a flood of fake celebrity profiles and brand impersonations worldwide.

    Despite the increase in online fraud, Meta’s internal assessment from 2022 described scam ads as a “low severity” issue, mainly considering them a poor user experience. Focus was primarily on impersonation scams involving well-known figures and brands, which posed a threat to advertiser relations and revenue.

    Layoffs and resource reallocations between 2022 and 2023 hindered enforcement efforts. Staff focused on brand concerns were let go, with many safety team resources diverted toward virtual reality and AI projects. As a result, the company largely neglected to address the growing tide of scam activity, with some internal documents revealing that up to 96% of user-reported scams were ignored or mishandled.

    Meta’s safety team aimed to improve this response, intending to process a higher percentage of valid scam reports moving forward. For example, in 2023, the team aimed to lower the rejection rate of legitimate scam reports from 96% to no more than 75%.

    Erin West, a former prosecutor now leading a nonprofit against online scams, criticized Meta’s general approach of ignoring user reports, noting that genuine reports often go unaddressed.

    An incident involving a Canadian military recruiter’s Facebook account highlighted these issues. After the account was hacked and impersonated, Meta was slow to respond, and the scammer used the profile to promote fake crypto investments. Despite multiple reports from the user and her contacts, Meta took about a month to take the account offline. This delay resulted in other military colleagues falling victim to scams, with some victims losing thousands of dollars.

    Authorities struggled to recover the stolen funds, which had been transferred to bank accounts in Nigeria, out of reach for enforcement due to jurisdictional restrictions. Meta declined to comment specifically on this case.

    Meta classifies scams into “organic,” meaning unpaid fraudulent activities like fake marketplace listings or charlatan health groups, as well as “paid” scam ads. The company estimates it faces about 22 billion organic scam attempts daily, on top of the 15 billion paid scam ads shown to users each day.

    The platform’s current policies often miss much of this activity. For example, Singaporean police provided Meta with 146 scam examples last fall, only to find that less than a quarter violated the site’s policies, with some violations falling short of the letter of the rules but still representing deceptive practices.

    Other flagged scams, like crypto ads claiming to feature the Canadian prime minister, were not caught under existing policies, although Meta later removed them for other reasons. Internal documents acknowledge significant gaps in the platform’s scam detection policies, highlighting the challenge in closing these loopholes.

    Even when advertisers are caught, enforcement can be lenient. A 2024 report states that small advertisers engaged in financial fraud typically have to be flagged multiple times—up to eight—before they’re shut down. Larger accounts, known as “High Value Accounts,” may evade restrictions after more than 500 violations.

    Some anti-scam campaigns have reached enormous sizes, with four campaigns alone responsible for $67 million a month in ad revenue before being removed. To further combat scams, Meta has experimented with increasing the costs for suspected rogue advertisers through “penalty bids,” which require high-risk advertisers to bid more to place ads. This strategy aims to make scams less profitable and reduce the number of scam ads shown.

    Following the implementation of this system, Meta reported declines in both scam reports and overall ad revenue, with the company aiming to strike a balance between financial interests and platform safety.

  • Google to discontinue Dark Web monitoring service

    Google to discontinue Dark Web monitoring service

    In 2023, a service designed to track leaked personal data was launched but will be discontinued early next year. Google is planning to retire its Dark Web Report tool, which was created to notify users when their personal information appears in illicit data dumps on the dark web.

    Google announced that the dark web data scanning feature will cease on January 15, 2026, with complete shutdown scheduled for February 16, 2026. All user data associated with this service will be permanently deleted from Google’s servers.

    The Dark Web Report was introduced in March 2023 as part of Google’s initiative to help users monitor their online footprints. It scanned for email addresses and other personal details such as names, phone numbers, and addresses to see if they appeared in known data breaches or dark web listings, alerting users if matches were found.

    In an email, Google stated, “While the report offered general information, feedback indicated it didn’t provide helpful next steps. We’re making this change to focus on tools that deliver clearer, actionable steps to protect your information online. We will continue to monitor and defend against online threats, including those on the dark web, and develop tools to safeguard you and your personal data.”

    Although the Dark Web Report service is ending, Google reassures users that efforts to protect their information will continue behind the scenes. The company recommends leveraging other security tools, such as Security Checkup, Passkeys, Password Manager, two-step verification, and Google Authenticator, to enhance account security. Additionally, features like “Results about you” assist users in locating and sometimes removing personal information from search results.

    Existing users of dark web monitoring are informed via email and support pages that access will end after the service shutdown. Users who wish to remove their monitoring profiles sooner can do so through their account settings by navigating to the Dark Web Report page, selecting “Edit monitoring profile,” and then “Delete monitoring profile.”

    To manually delete your dark web monitoring profile early, go to your account, click on the Dark Web Report page, select “Edit monitoring profile,” and then choose “Delete monitoring profile.”

  • Students Reveal Innovations at Karachi Sci-Tech Festival

    Students Reveal Innovations at Karachi Sci-Tech Festival

    The three-day Sindh Youth Science and Technology Festival is scheduled from November 28th to 30th at the Sindh Youth Club in Karachi.

    Karachi, Sindh:
    Students from across Sindh have gathered in Karachi to display their innovative projects at the three-day Youth Science and Technology Festival, which features everything from robots and green energy solutions to insects capable of breaking down garbage.

    The event was inaugurated on Friday by Sardar Muhammad Bakhsh Mahar, the Minister for Sports and Youth Affairs. Over 70 students representing around 40 institutions are presenting their projects, with prize money exceeding 200,000 rupees allocated for the top entries—100,000 rupees for first place.

    Mahar explained that a panel of professors, specialists, and industry professionals has been assembled to assess each project with fairness and technical rigor, ensuring transparent evaluation and high standards.

    During my conversations with students yesterday, nearly everyone mentioned that their projects were the result of months—sometimes up to a year—of research and hard work. The event hosts over 50 booths, with some students sharing their displays. While it’s impossible to detail every project, here are some of the most innovative ones on display.

    Garbage-Disposal Larvae

    A team from SST Public School in Rashidabad presented black soldier fly larvae and mealworms designed to decompose waste at minimal cost, according to students Ravi and Sidra.

    “These are zero-cost,” explained Dr. Maleeha Jamil, noting that these insects produce numerous offspring simultaneously, making them an effective waste management method. She added that the larvae and mealworms can consume biowaste and even plastics, though only biowaste was shown during the demonstration.

    I recorded a video of the pupae and larvae—viewer discretion advised for those squeamish about insects.

    Many might assume that feeding so many larvae would attract swarms of beetles and flies around waste sites. However, the students have a solution: “These larvae are about 60% protein,” Sidra said, suggesting they could serve as animal feed. Experiments adding them to chicken and fish feed resulted in larger eggs and bigger fish with higher Omega-3 levels.

    Further testing is needed to understand any possible health impacts or unforeseen side effects, as well as to quantify the actual benefits of incorporating these larvae into animal diets.

    What struck me most was Ravi’s remark that “about 1 kilogram of larvae can eat approximately 3 kilograms of waste,” which immediately made me think of the pseudo-landfill near my house—I almost wanted to ask them for a handful right then.

    Codion: The Bilingual Talking Robot

    Humbal and Barirah developed Codion, an AI-powered humanoid robot capable of speech, movement, and interactive responses. Unlike the robots from big companies like Meta or Boston Dynamics, this one is a simple walking skeleton on wheels.

    The robot can hold conversations, follow commands to walk or roll, speak in both English and Urdu, share information about itself, and, if connected to the internet, perform online lookups.

    It can recite Urdu poetry, although I couldn’t test this due to slow internet at the venue. The programmers explained that the robot has onboard memory to answer questions, but Humbal admitted it doesn’t “know” that it’s part of a demonstration—a sign of its early development stage.

    Wall-E’s Prototype Counterpart

    Inspired by the beloved childhood film, this robot was created by Simrah, a student from Karachi’s Shaheed-e-Millat College for Women. It navigates using a built-in camera feed and has a compartment for large parcels on its front. The goal is to develop a robot capable of transporting, serving, and delivering items in places like airports, malls, and restaurants.

    While not fully autonomous yet, it can be remotely controlled and adjust its grip and movement. However, it struggles on carpets, which the team is actively working to improve.

    Autonomous Shopping Cart

    From Mehran University in Khairpur, Peeryo designed a shopping cart that can follow a shopper, scan item prices, sense weight, and count the items in the cart.

    “This could be a huge help for people with disabilities,” he said while demonstrating the cart’s functionality via a custom app. I tested it myself—though slow— and it managed to follow me, but it lost sight when I moved too far away.

    The prototype is small and can’t carry large loads yet, but future designs will be larger to accommodate more.

    Solar-Powered Wind Turbine

    Yasir, Muhammad, and Arslan from the University of Larkana showcased a turbine with three flexible solar panels shaped into a helix. The model was assembled locally, with the blades imported from China.

    “Combining solar and wind energy, this system can generate power simultaneously,” Yasir explained. The turbine’s unique shape allows wind to turn it from multiple angles and maximizes sunlight exposure on the panels.

    Sensors measure power output, and the design allows connection to the grid or batteries. Yasir added that the near-constant movement and helix shape help protect photovoltaic cells from dust damage. The design is scalable for various sizes—suitable for rooftops or energy parks.

    Biowaste Burning for Electricity

    From Mehran University, Salman, Faheem, and Raheel shared a model that uses photovoltaic thermal (PVT) cells to capture energy from burning agricultural waste.

    They explained that Pakistan generates around 40 million tons of agricultural waste annually, which, if burned for energy, could produce up to 15,000 MW. They proposed a plant could power around 100,000 households for three decades, using a one-time investment of approximately $1.2 million.

    However, burning large quantities of biological waste can harm the environment, raising concerns about short-term benefits versus long-term ecological impacts.

    Biodegradable Bandages

    Nimrah, Saira, Nisha, and Sufiya from the Jinnah University for Women developed plant-based, biodegradable bandages to reduce plastic waste in marine environments. Using starch for bioplastic, okra gelatin as adhesive, and pomegranate peels as antiseptics, their design aims to produce sustainable wound coverings.

    Portable Power Supply for Emergency Medical Use

    Ijaz and Abdul from Indus University created a portable power system for medical equipment in ambulances, which they presented to Sindh Ambulance Rescue services.

    “This system is vital if someone has a heart attack in an ambulance stuck due to heavy rain or in remote areas with no stable power,” Ijaz explained. The device can run one medical machine—like a respirator or monitor—for 18-20 hours, powered by a lithium iron phosphate battery that’s safer than lithium-ion.

    It can be recharged in 1.5 hours via solar or 15 minutes from main power and weighs 13.5 kg, with transportable options like a suitcase. The system features separate control modules for user operation and remote management.

    SupeRAFT: The Marine Rescue Drone

    Moiz from Sir Syed University unveiled SupeRAFT, a remote-controlled rescue device capable of carrying up to 200 kg. With a 1-hour battery life, it’s designed to quickly respond to drowning victims at sea, especially where deploying human rescuers might be risky.

    It has dual propulsion, radar, computer vision for precise location tracking, and underwater scanning capabilities. The prototype cost between 200,000 and 300,000 rupees.

    More Innovations and Talents

    The event showcased numerous other impressive projects. Indus University displayed AI-powered goggles assisting visually impaired individuals, while a team from Hamdard University introduced AI glasses paired with specialized keyboards for users with neuromuscular conditions.

    Ayesha from Dow University presented dental tablets alternative to toothpaste, aiming to reduce millions of plastic tubes landfilled annually. Lahin demonstrated Velcro-based bandages designed specifically for foot ulcers common in diabetics.

    This event highlights the extraordinary talent among Pakistan’s youth. The awards ceremony will be on Sunday, recognizing the top three teams among many more deserving projects.

  • Karachi Residents Experience Internet and Service Outages

    Karachi Residents Experience Internet and Service Outages

    A screenshot of Downdetector’s page on November 7, 2025. Source: The Express Tribune

    Last night around 10 p.m. on Thursday, residents across Karachi experienced widespread disruptions to their internet and online services. Providers like PTCL and Stormfiber, along with platforms such as WhatsApp, YouTube, Amazon Web Services, and Instagram, reported outages, according to Downdetector.

    The outage monitor indicated that both PTCL and Stormfiber encountered problems between 9 p.m. and midnight. PTCL registered a peak of 112 outage reports around 10:30 p.m., while Stormfiber had 59 reports. Most users complained about losing internet access, though some faced complete service blackouts.

    Read: Faisalabad Implements E-Challan System

    On Friday, a PTCL spokesperson denied any outages occurred. “There was no issue on PTCL’s end,” they stated.

    Meanwhile, Stormfiber’s automated response stated, “We are experiencing a sudden degradation with one of our upstream providers,” and assured customers that they are working with the upstream provider to resolve the issues.

    Images showing user-reported outages over the past 24 hours reveal a sharp spike last night. The first graph illustrates PTCL outages, and the second shows Stormfiber outages, both with notable increases.

    Social media platforms also went down. Reports of Instagram and Facebook outages peaked at 134 and 65, respectively. YouTube and X experienced outages, but with fewer reports—16 for YouTube and four for X.

    A tweet from November 6, 2025, asked, “Can someone tell me if it’s my PTCL connection dying or if Twitter is lagging in Pakistan again? 😭😭”

    WhatsApp outages peaked with 120 reports, with 55% of users struggling to send messages, 32% facing server connection issues, and 13% experiencing app problems.

    Overall, most complaints centered around website access, app functionality, and server connectivity, with the majority of users reporting an inability to connect to the internet.

    Many Stormfiber users reported having very poor internet last night; however, one individual noted little change in their connection. Most issues lasted until midnight, with service levels returning to normal by this morning.

  • Boxed Smartphone Imports in Pakistan: Choice or Protection?

    Boxed Smartphone Imports in Pakistan: Choice or Protection?

    Sealed Pixels, Nothing, and flagship gaming devices are quickly becoming available, but buyers face challenges like PTA fees and uncertain warranties.

    A growing number of Pakistani consumers are purchasing smartphones outside official local launches. Sealed Google Pixels, Nothing phones, OnePlus flagships, ROG and Redmagic gaming devices, and boxed iPhones are being sold through parallel imports and specialized resellers. This provides faster access to niche models but also increases the risk of regulatory issues, warranty problems, and repair uncertainties.

    These devices generally arrive factory sealed. Sellers often provide shop-backed warranties and PTA registration services; however, the level of official support varies by brand and batch.

    Sources: Pakistan Telecommunication Authority (DIRBS FAQ, IMEI registration rules); Federal Board of Revenue mobile duty schedules; market prices from Yellostone Pakistan, Cube Online, PhoneDroid, OLX listings, Hafeez Centre (Lahore)

    Impact of the grey imports market on consumers

    The grey import economy allows buyers to access desired models quickly, sometimes even before they are officially launched in Pakistan. However, this convenience comes at a price. Final costs often increase once PTA registration, brokerage fees, and customs duties are factored in, meaning the out-the-door price can be significantly higher than the advertised price. Warranties offered on these units are typically provided by shops rather than manufacturers, limiting recourse in case of hardware or software issues.

    Repair times tend to be longer because region-specific spare parts are not typically available locally and must be imported. Additionally, some models require firmware or region adjustments that can prevent over-the-air updates or restrict certain network features. While buyers benefit from early access and novelty, they also face hidden costs, longer downtime, and uncertain long-term support.

    Understanding the PTA tax gap explains why sealed Pixel and OnePlus prices fluctuate widely.

    Listings for imported phones usually separate non-PTA stock from PTA-approved batches. Non-PTA devices may be cheaper initially, but buyers often pay extra for IMEI registration, customs duties, and brokerage fees. Retailers often market PTA approval as an optional add-on or promote limited PTA-cleared stock at higher prices. This creates a confusing price disparity; for some flagship models, the difference can amount to tens of thousands of rupees.

    Where repairs are done (and why this is risky)

    Most parallel imports lack authorized manufacturer service centers in Pakistan. Repairs are consequently handled by independent technicians or the selling shops. Region-specific variants—like some Pixels and gaming models—often require parts ordered individually, extending repair times and raising costs. While seller warranties exist, they are contractual promises from retailers rather than official manufacturer guarantees, and their enforcement and spare part availability differ depending on location and seller.

    How retailers market ‘global variants’ and ‘PTA approved’ phones as premium options

    Stores often label phones as “global ROM,” “US box,” “factory unlocked,” or “PTA ready” to justify higher prices for certain stock. Influencer reviews and international unboxing videos boost demand for these variants, which sellers capitalize on with upsells such as IMEI verification, PTA paperwork handling, and extended shop warranties. While these measures facilitate smoother transactions, they do not substitute for manufacturer-backed after-sales support.

    Position of iPhones in the market

    iPhones are available through both channels: PTA-approved boxed units sold via authorized resellers and cheaper grey imports listed on classifieds and in street markets. Global warranty coverage is limited without Apple’s official service presence in Pakistan; consumers often rely on reseller warranties or third-party repair networks. Persistent rumors suggest a possible Apple authorized reseller arrangement, which, if confirmed, could clarify warranty and service concerns.

    Example: Nothing’s efforts toward formalization

    The market’s structure is evolving. In an interview with The Express Tribune, Yellostone confirmed it is the official distributor for select Nothing Phone models in Pakistan, offering local warranty support. However, this formal channel is limited to specific models, and many other Nothing variants, along with competing brands, continue to be primarily available through parallel importers.

    A balanced approach for buyers

    If you’re interested in importing or early models, consider the following steps:

    1. Check IMEI status on the PTA DIRBS portal prior to purchase.

    2. Obtain a written warranty from the seller or distributor, detailing coverage duration and what faults are covered.

    3. Confirm availability of spare parts and typical repair timelines with the seller.

    4. Compare the final price—including PTA registration—rather than just the initial sticker price.

    5. Prefer units sold through recognized distributors (like Yellostone for certain Nothing models), especially if warranty and parts support are priorities.

    Today’s boxed imports are an integral part of Pakistan’s smartphone landscape. They offer consumers options and early access but shift post-sale responsibilities onto buyers. Whether this parallel import channel remains a temporary workaround or becomes a permanent feature depends on clearer regulatory policies, expanded official distribution, and stricter consumer protection enforcement.

  • Pinterest Allows Users to Disable AI-Generated Content in Feed

    Pinterest Allows Users to Disable AI-Generated Content in Feed

    AI-generated content is widespread these days, making it almost impossible to avoid. With tools like Veo 3 and Sora 2 for video creation, and Gemini’s Nano Banana for image generation, anyone with a subscription can produce professional-quality content within minutes.

    Luckily, Pinterest has introduced a feature that allows users to block AI-generated images from appearing in their feeds. Many users, including popular illustrator @hansoeii on X, are celebrating this update as a positive step forward.

    If you’re a Pinterest user and want to disable these images, here’s what you need to do:

    How to disable AI-generated images

    1. Sign into your account
    2. Access your settings
    3. Select the “Refine your recommendations” option
    4. Navigate to the “GenAI interests” tab
    5. Turn off all options in that section

    When tested with a few colleagues, everyone was surprised at how simple it was to turn off the feature. It’s a welcome respite, especially as companies like Meta continue to prioritize AI content in our feeds.

    However, this feature isn’t flawless. Users like @yogeshgosavi have reported that their feeds still show AI content despite turning off the option. Others have mentioned that disabling AI images negatively impacts the relevance of their recommendations.

    Additionally, turning off AI content doesn’t entirely eliminate it. When you disable an option, a pop-up warns that doing so will only reduce the amount of AI-generated content, not remove it completely.

    Furthermore, Pinterest updated its privacy policies earlier this year, stating it may collect and use your data to train and refine its machine learning and AI models, regardless of when Pins were posted. This means your data continues to influence AI training, even if AI content is less visible in your feed.

    While Pinterest cannot fully separate itself from GenAI, many artists and illustrators see this move as a positive direction. Unlike Meta, which is pushing the integration of AI across various platforms — from data collection to personalized feeds — Pinterest appears to be slowing down the AI rush, giving users more control over their browsing experience.

  • Surveillance States, Self-Driving Cars, and AI Doomsaying: The Norm

    Surveillance States, Self-Driving Cars, and AI Doomsaying: The Norm

    They want surveillance states to become the new normal

    Source: Reuters

    ICE plans to hire 30 contractors whose sole responsibility will be monitoring social media for posts that could lead to deportation actions.

    They are seeking private vendors to run a multi-year surveillance program, scanning platforms like Facebook, TikTok, Instagram, YouTube, and others, using user posts and profiles as leads for enforcement raids.

    Although still in early development, draft planning documents reveal a focus on 24/7 surveillance, rapid case processing under tight deadlines, and the allocation of funds for advanced surveillance software. There are also discussions about integrating AI into this system.

    This discussion revolves around ICE, but it reflects a broader global trend. Our data is already being sold to the highest bidder. Recently, Meta announced they will use your personal chat data with their AI starting in December—without an option to opt out. This move exemplifies how companies are increasingly disregarding privacy and encroaching on our digital personal space, threatening our security.

    Read: The risks of iRobot, privacy concerns with Meta, and Nintendo’s Lego Game Boy project

    These issues won’t just go away on their own. Maybe it’s already too late to stop their escalation, but perhaps we can prevent them from worsening. Don’t accept the new normal—resist it.

    Abandoned cities and empty vehicles

    Source: ReutersSource: Reuters

    In Ordos, China, companies specializing in self-driving cars are actively testing their vehicles.

    After the 2012 real estate crash, Ordos experienced a significant population decline, with over 70% of buildings remaining unfinished due to economic downturn. Its wealth from coal exports fuels demand for cargo logistics.

    Currently, over a dozen autonomous vehicle firms have set up shop here, offering transportation services while testing their self-driving tech.

    While the sparse environment provides a safe testing ground, it doesn’t reveal how these vehicles perform in crowded urban settings.

    It’s reassuring that testing occurs in controlled areas, but I wouldn’t want these trucks sharing the same roads as pedestrians and city traffic without more comprehensive testing in real city conditions.

    Business and AI: a digital power couple

    Source: Google GeminiSource: Google Gemini

    AI has deeply ingrained itself within corporate life.

    A recent study published in the journal Patterns found that in corporate, government, and media sectors, AI tools are routinely used to draft everything from job postings to press releases.

    Even the United Nations has been using AI to prepare content over the past couple of years. The study indicates that the proportion of text likely generated by AI increased from 3.1% at the start of 2023 to 10.1% by the third quarter, peaking at around 13.7% a year later.

    On LinkedIn, about 6-10% of sampled job listings contained some form of AI writing. Smaller companies tend to use AI more frequently than larger ones.

    While AI’s role in press releases has stabilized around 24.3% since December 2023, AI-generated content continues to fill the digital landscape at a rapid pace.

    There may soon be hardly anything online that isn’t at least partly AI-created, and this isn’t a distant future—it’s happening right now.

  • Deep Ocean Data Centers, AI Toxins, and the Memory Crisis

    Deep Ocean Data Centers, AI Toxins, and the Memory Crisis

    China’s Deep Dive into Data—Quite Literally

    Underwater data center under construction at a shipyard in Nantong, eastern Jiangsu province, developed by Chinese maritime tech firm Highlander. Source: x.com/japantimes

    China has truly taken this idea to a new level.

    They’re building a server pod near Shanghai that will be submerged underwater to cut cooling costs by about 90%, they claim.

    Highlander, the Chinese company behind this project, plans to sink the servers later this month. Once operational, China Telecom is expected to be among the first customers.

    Of course, saltwater and electronics don’t play well together, so extensive safety measures had to be implemented before this venture could proceed.

    While underwater data centers aren’t entirely new—Microsoft tested a similar setup off the coast of Scotland in 2024, and China has been involved in an underwater project since 2023 in Hainan—this marks an improvement. Building data centers on land is costly due to cooling expenses, but Highlander claims 95% of their energy is sourced from renewable resources.

    Water and electricity are the key challenges when constructing data centers. If more nations and companies adopt this approach, it could benefit the environment and reduce operational costs. However, experts are worried about ocean warming from the continuous heat output of data centers.

    Further research is essential to weigh the benefits against potential environmental tradeoffs.

    Biological Warfare 2.0

    Molecular structure of the protein myosin. Source: x.com/NikoMcCartyStructure of the protein myosin. Source: x.com/NikoMcCarty

    A team of researchers led by Microsoft uncovered a “biological zero day”—an undetected security flaw in DNA screening systems. AI can now design unique toxins that evade current biosecurity checks because they don’t match anything known.

    DNA synthesis companies typically scan their transcribed sequences against databases of known dangerous substances. If anything appears suspicious, human experts step in. This method has been in place for years, with governments collaborating closely with industries to prevent new threats, expanding screening from simple sequences to larger proteins.

    Problems arise because multiple DNA sequences can encode the same protein, and different proteins can perform similar functions, including toxicity. After testing various screening methods, researchers found that only 1-3% of out of 75,000 variants of 72 toxins went undetected.

    This is somewhat reassuring—AI tends to generate non-functional proteins more often than dangerous ones. Overall, weaponizing this technology remains difficult, because an attacker would need to order numerous sequences to find one that both acts as a toxin and bypasses screening. Yet, as AI tools improve, they may increasingly produce toxins that are both effective and undetectable at higher rates.

    To counter this threat, developing advanced biosecurity software and more sophisticated screening protocols is critical.

    The Coming Crisis in Memory and Data Storage

    Source: ReutersSource: Reuters

    “Today, I’ll set up thousands of data centers that demand massive resources to build and operate. Surely, this won’t lead to any unforeseen global issues.”

    First, AI took our jobs. Then, data centers drove up your electricity bills. Now, there’s talk of underwater data centers and AI-generated toxic proteins that could potentially be weaponized.

    The good news keeps getting trickier to find.

    The worldwide memory chip market—including NAND, DRAM, and HDDs—is heading toward a severe supply shortage that could last a decade, fueled by the skyrocketing demand for AI data centers and the slow pace of production growth.

    Prices for SSDs, RAM, and even hard drives are climbing again. Large firms are pre-purchasing future supplies of memory chips years in advance to support the training of large language models and more advanced AI systems, which is only tightening supply for everyone else.

    Meanwhile, manufacturers are cutting back on investments in NAND and DRAM production, which take years and billions of dollars to develop.

    Trade disputes and tariff wars are further constraining supply chains, exacerbating the shortage.

    Experts warn that the AI industry might be experiencing another bubble, similar to the dot-com era. Much of the current investment might not pay off, and investors are growing more cautious about sinking money into AI projects.

    Read: The Age of Uncertainty: AI’s Impact on Jobs and Policy

    If you’re in the market for a high-performance PC—whether for work or gaming—it’s probably best to buy sooner rather than later, as prices are unlikely to drop anytime soon.