Tech Roundup: OpenAI Lawsuits, Teen Social Media Ban & More

[A recurring feature on the latest in Science & Technology.]

• Oura launches a redesigned app with a new feature showing a weekly overview of stress management and says it's working with the U.S. FDA on a blood pressure study.
• Adobe launches an AI Foundry program that helps enterprise customers create bespoke, commercially safe, Firefly-based generative AI models trained on their intellectual property.
• The U.K. antitrust regulator says Getty Images and Shutterstock's US$ 3.7 billion merger raises competition concerns and gives the companies until October 27 to make a plan to address them.
• Spotify adds a new feature that lets users follow their favourite concert venues as a way to provide updates and details about upcoming concerts; introduces a new "listening stats' feature that gives users a weekly overview of their recent music streaming habits, following a similar move from Apple, which debuted a monthly version of Apple Music's annual Replay feature in February 2024.
• DeepSeek releases DeepSeek-OCR, a vision language model designed for efficient vision-text compression, enabling longer contexts with less compute.
• Ticketing giant Ticketmaster says it will ban users from operating multiple accounts, shut down its TradeDesk ticket inventory platform and start requiring ticket brokers to hand over their Social Security numbers in order to sell tickets on its resale platform following antitrust concerns.
• TikTok says it will inform users only where required by law and changes the timing from before disclosure to if disclosure occurs. (The company's policy, until April 25, 2025, stated that it would notify users before disclosing their data to law enforcement.)
• Google launches new features for its Google AI Studio developer platform, including a revamped "Build" tab to vibe code AI apps; plans to add an AI call quality feature in Google Fi to filter out background sounds like wind next month, and bring full RCS support on the web in December 2025.
• OpenAI launches its ChatGPT Atlas web browser, available initially for macOS, as it takes on Chrome and other agentic AI browsers like Dia, Perplexity Comet and Opera Neon; features an agent mode to let ChatGPT automate web-based tasks on behalf of ChatGPT Plus and Pro users and includes an opt-in browser memories feature that can remember key details from users' web browsing to improve chat responses and offer suggestions.
• Samsung rolls out a Perplexity TV app, which works alongside Samsung's Vision AI Companion, on its 2025 TVs.
• Google's YouTube launches its likeness detection tech, letting eligible creators in its Partner Program request the removal of AI-generated content with their likeness to combat deepfakes; adds a Shorts feature that lets users setting a scrolling time limit on the Shorts feed and tests custom like button animations for different types of videos.
• Airbnb adds direct messaging, new social features following the travel company's major overhaul in May 2025, which brought services such as catering and personal training to its app.
• Russia pressures Apple to make Russian search engines such as Yandex or Mail.ru the default option on locally-sold iPhones; gives the company time till October 31, 2025, to comply with the authorities' demands or face penalties.
• Chinese internet search and artificial intelligence giant Baidu teams up with Swiss public transport operator PostBus to bring robotaxi services to Switzerland.
• Apple confirms it will not be allowing customers to trade in the original Vision Pro toward the purchase of the new model or any other device.
• The U.K. Competition and Markets Authority (CMA) confirms that both Apple and Google have strategic market status (SMS) in each of their mobile platforms, adding the companies "have substantial, entrenched market power and a position of strategic significance in their respective mobile platforms."
• Russia tightens control over its digital space by promoting MAX, a state-run messaging "super app," while restricting voice and video calls on WhatsApp and Telegram, citing "antifraud" efforts; critics see it as a push toward a government-monitored internet.
• Tinder plans to expand its facial verification feature to all new U.S. users in the coming months, following its rollout in seven countries and in California.
• Apple removes controversial dating safety apps Tea and TeaOnHer from the App Store for failing to meet requirements around content moderation and user privacy; the apps continue to be available on Google Play.
• Reddit files a lawsuit against Perplexity and three data scraping companies SerApi, OxyLabs and AWMProxy, accusing them of illegally stealing its data by scraping Google search results.
• Google says its Willow quantum chip using a new Quantum Echoes algorithm called Out-of-Time-Order Correlator (OTOC) runs computations 13,000x faster than supercomputers, demonstrating a verifiable quantum advantage and aiding drug and materials research.
• GM plans to roll out Google Gemini in its vehicles starting in 2026 and plans an advanced driver-assistance system with eyes-off driving capabilities in 2028.
• Samsung unveils Galaxy XR headset with Android XR, 4K micro-OLED displays, a Snapdragon XR2+ Gen 2 chip and an external battery pack; becomes the first Android XR headset and features Gemini Live to understands users' surroundings and lets them interact with apps hands-free. (It also comes with an Explorer Pack that bundles a year of Google AI Pro, YouTube Premium, Play Pass and more.)
• Amazon introduces "Blue Jay" warehouse robot to lift, sort and consolidate packages and help reduce "cognitive load"; unveils AI smart glasses for its delivery drivers to find the right packages inside their delivery vans, give them turn-by-turn directions to the right address and take a hands-free photo of a successful delivery.
• Snapchat expands access to its AI-powered Imagine Lens, allowing all users to generate and edit images with custom prompts for free.
• Meta's Instagram debuts custom icons for its app, but only for teen users, and says it has no plans to expand access to icons to older users or paid subscribers; adds new watch history for Reels and generative AI-powered photo and video editing tools to Stories, allowing users to use text prompts to remove or change objects in your photos or "restyle" the image.
• The U.K.'s Competition Appeal Tribunal rules that Apple abused its dominant position by charging app developers unfair commissions through its App Store, potentially costing the company hundreds of millions in damages; marks the first major tech "class action" victory under the U.K.'s collective lawsuit regime.
• Microsoft says it won't provide erotica AI services and that "other companies will build that"; updates Copilot to add Copilot Groups, enabling up to 32 people to collaborate in a session, and a "real talk" mode to add more personality to Copilot, and unveils Mico, a character that responds with real-time expressions when talked to, in Copilot's voice mode in the U.S., U.K. and Canada.
• Amazon blames a rare software bug and "faulty automation" for a massive AWS glitch that crippled several sites and online services around the world, and says it has turned off the flawed automation worldwide.
• Anthropic and Google announce cloud partnership worth tens of billions of dollars, giving Anthropic access to 1 million TPUs and 1GW of capacity in 2026; comes as Apple's Houston manufacturing facility, originally scheduled to open in 2026, begins shipping AI servers that power Apple Intelligence and Private Cloud Compute.
• Anthropic says Claude's memory feature, initially available for Team and Enterprise users, is rolling out to Pro and Max subscribers; expands Claude for Financial Services with a beta Claude for Excel integration, additional data connectors and new pre-built Agent Skills.
• OpenAI acquires Software Applications Incorporated, a startup building an AI-powered user interface for macOS (The company is best known for making Workflows, which was later purchased by Apple in 2017 and rebranded to Shortcuts); updates Sora, its hit AI video app, with new features, including expanded character cameos, video editing tools, leaderboards to highlight the most popular videos and cameos, and enhanced social options, and launches company knowledge in ChatGPT, letting users access their firm's data from Google Drive, Slack and GitHub.
• Amazon relaunches its game streaming service Luna, available at no additional cost for Prime subscribers, and adds new beginner-friendly multiplayer games that it says are "designed to bring friends and family together in the living room"; unveils a new AI feature in its namesake shopping app called Help me decide, which takes into account user searches, browsing and shopping history to suggest new products.
• Dropbox integrates some of Dash's AI features into its main app, giving its users access to a smarter search function, summaries and contextual answers from the files they've uploaded to the cloud service.
• Microsoft officially launches a Copilot Mode in Edge, after first testing the feature in July 2025; offers the ability to start a chat, search or navigate the web, reason across multiple open tabs, complete multi-step actions, and provide richer insights using users' past browsing history (if the option is enabled).
• Google updates Google Earth with Gemini capabilities to find objects and discover patterns from satellite imagery; comes as the company introduced Google Earth AI, its collection of geospatial AI models and datasets, to help improve weather predictions, flood forecasting, wildfire detection, urban planning and public health.
• A group of academics from Texas A&M, the University of Texas, and Purdue University find that training on "junk data" can lead to LLM "brain rot," warning "heavily relying on Internet data leads LLM pre-training to the trap of content contamination."
• SpaceX says it shuts down thousands of Starlink terminals powering Myanmar's notorious scam compounds, marking one of the clearest acknowledgements yet from the company that its low Earth orbit broadband service has been exploited by criminal groups.
• The European Commission accuses tech giants TikTok and Meta of breaching their obligation to give researchers "adequate access" to public data under the Digital Services Act; calls out Meta for failing to give users easy ways to flag illegal content and adequate tools to appeal moderation decisions.
• Microsoft plans to roll out a new feature in Teams that automatically sets users' work location "to reflect the building they are working in" when connecting to their organization's Wi-Fi; comes as Google adds support for waiting rooms in Meet.
• OpenAI launches a "company knowledge" ChatGPT update to let Business, Enterprise, and Education users search for data in connected apps like Slack and GitHub; estimates that around 0.07% of ChatGPT users active in a week show "severe mental health symptoms" like mania and that 0.15% of its weekly active users have "conversations that include explicit indicators of potential suicidal planning or intent."
• India's government proposes that artificial intelligence and social media firms should clearly label AI-generated content to tackle the spread of deepfake and misinformation, prompted by similar moves by the European Union and China; says the potential misuse of generative AI tools "to cause user harm, spread misinformation, manipulate elections, or impersonate individuals has grown significantly."
• Australia sues Microsoft for allegedly misleading about 2.7 million Microsoft 365 users into accepting a price hike for Copilot plans while hiding cheaper non-AI options.
• Microsoft says its Gaming Copilot feature, which recently hit Windows 11 as a public beta, uses screenshots to understand in-game events, not to train AI models.
• Google begins rolling new Gemini Canvas feature that lets users create presentations using text-based prompts and export them to Google Slides; updates its Chrome browser with capabilities to fill in users' passport, driver's license, vehicle registration and more as part of its auto-fill feature.
• X plans to retire the Twitter.com domain, prompting users to re-enroll their security keys for 2FA; to lock accounts that are not updated by November 10.
• Apple says it will roll out a new Apple Wallet feature that lets U.S. users create digital IDs linked to their passports.
• Grokipedia, a Grok-powered online encyclopaedia, launches with over 885,000 articles, as it claims to be a "less biased" alternative to Wikipedia. (According to WIRED, the new AI-powered Wikipedia competitor falsely claims that pornography worsened the AIDS epidemic and that social media may be fuelling a rise in transgender people.)
• Apple scores a legal victory after a U.S. court decertifies a class action ruling in February 2024 that accused Apple of monopolising the market for iPhone apps, citing errors in the plaintiffs' damages model; comes as a group of 55 Chinese iPhone and iPad users filed a complaint with China's market regulator, the State Administration for Market Regulation, alleging that Apple (AAPL.O), opens new tab abuses its market dominance by restricting app distribution and payments to its own platforms while charging high commissions.
• Microsoft is reportedly planning a "premium" experience for its next generation Xbox platform; to run Windows, letting users play PlayStation titles on Steam and the full Xbox console library, without a multiplayer paywall.
• China's MiniMax releases MiniMax M2, an open-source model optimised for AI agents and coding.
• Meta launches "ghost posts" on Threads, allowing users to share "unfiltered thoughts" in posts that disappear after 24 hours, with replies to the posts appearing as a DM; adds new controls that lets users filter replies in its Activity tab and approve replies to their posts.
• IBM launches Digital Asset Haven, a digital assets platform built with crypto wallet provider Dfns for financial institutions, governments and companies.
• Spotify brings video podcasts and music videos to its redesigned Apple TV app; also integrates the ability to manage queue and view lyrics.
• Samsung rolls out a new update for Family Hub refrigerators that includes a refreshed One UI design, smarter food tracking with AI-powered technology, access their calendar or photos, and find a misplaced phone; to also come with a new widget for select Cover screens themes that displays day-to-day information along with curated advertisements that can be dismissed. (While there is an option to turn off the ads, the development, coupled with recent appearance of full-screen ads on Amazon's Echo Show smart display, signals that any device with an internet connection and a screen is no fair game for ads.)
• Google's Gemini-powered health coach becomes available in Fitbit to Premium subscribers in the U.S. on Android, with plans to expand to iOS soon; develops a new framework called Watch and Learn (W&L), to gather high-quality training examples at scale and help train or fine-tune existing computer use and foundation models to improve their performance on computer-use tasks.
• Pinterest debuts new AI-powered upgrades to its boards to make them more personalised to users by curating a blend of editorial input and AI-powered suggestions featuring trending styles, weekly outfit inspiration and shoppable content; to launch an AI-enabled shopping assistant to suggest new clothing recommendations based on their saved collections.
• A survey conducted by Meta finds that teens who report that Instagram regularly made them feel bad about their bodies saw significantly more "eating disorder adjacent content" than those who did not, according to a report by Reuters.
• The U.S. Digital Childhood Institute files a complaint with the U.S. Federal Trade Commission (FTC), accuses Apple and Google of engaging in "deceptive and unfair" business practices that violate the Children's Online Privacy Protection Act (COPPA); says Apple and Google have agreed to change their internal tools to "verify users' ages, require parental consent before minors can download or purchase apps, and share limited, privacy-preserving age data with developers."
• Meta and TikTok say they will comply with Australia's under-16 social media ban, taking effect on December 10, but warn the landmark law may be hard to enforce; comes as the Danish government reaches an agreement to implement a minimum age requirement of 15 years old on certain social media platforms. (That said, 13-year-olds can still go on social media with their parents' consent.)
• OpenAI offers its cheaper ChatGPT Go plan for free for one year to users in India starting November 4, 2025; does not say how long the offer would remain available.
• PayPal signs deal with OpenAI to become the first payments wallet in ChatGPT, allowing users to pay for items starting in 2026.
• Adobe unveils Project Moonlight, an AI agent on its Firefly platform designed to act as a creative director for social media campaigns via text prompts; partners with YouTube to offer a Create for YouTube Shorts feature in the Premiere mobile app, letting creators instantly publish video.
• Adobe plans to support custom AI models in Firefly, as it updates Firefly to generate soundtracks and speech, and adds an AI assistant to Photoshop and Express; unveils AI tools for Photoshop, Premiere Pro and Lightroom to automate photo edits, and says Firefly Image 5 can generate images in native 4MP resolution.
• Adobe showcases a new Corrective AI tool can alter the emotional tone and style of a voice-over, transforming it from flat to confident or even a whisper within seconds.
• Microsoft and OpenAI finalise a new agreement as part of which Microsoft gets a 27% stake in OpenAI and access to its AI models until 2032, including AGI-level models, with OpenAI buying Azure services worth US$ 250 billion.
• Alibaba's Qwen3-Max takes commanding lead over DeepSekk and U.S. counterparts in a live real-world real-money cryptocurrency trading competition called Alpha Arena, posting triple-digit gains in less than two weeks.
• Microsoft announces a new Microsoft 365 Copilot agent called App Builder that can help users create and deploy apps "in minutes" using text-based prompts; also launches Workflows agent to automate tasks like sending emails and reminders, managing calendars and sharing team updates.
• Palantir announces a partnership to use Nvidia's chips and software to help its customers speed up decision-making in complex fields such as logistics, as Uber plans a fleet of 100,000 autonomous vehicles powered by Nvidia's new Drive AGX Hyperion 10, starting in 2027.
• Celebrity video shoutout app Cameo sues OpenAI for trademark infringement, alleging Sora's "cameo" feature is likely to cause confusion and dilute its brand.
• YouTube announces plans to age-restrict more video game content showing "graphic violence" and stop creators from directing users to gambling sites involving digital goods; offers creators a way to upscale videos uploaded at between 240p and 720p to "HD" using AI and plans to offer 4K upscaling "in the near future" and plans to enforce new age-restrictions around online casino and graphic video game content starting on November 17, 2025.
• China claims it has completed the first phase of construction of the first underwater data center, which is powered by wind energy and cooled by seawater.
• Oppo launches €999 Find X9 and €1,299 Find X9 Pro in the U.K. and E.U., with 7,025mAh and 7,500mAh silicon-carbon batteries and a MediaTek Dimensity 9500 chip.
• Microsoft-owned GitHub plans to launch Agent HQ, a hub to let developers access third-party AI coding agents, including Claude and OpenAI's Codex, in the coming months; updates Visual Studio Code with Plan Mode for building step-by-step project approaches, MCP Registry integration, definable project rules via AGENTS.md.
• Social media platform Truth Social launches Truth Predict, a prediction market platform in partnership with Crypto.com to let users trade on major current events like elections.
• The University of Southern California sues Google in Texas federal court, alleging that Google Earth, Maps, and Google Street View violate its patents for overlaying 2D images onto 3D models.
• Google begins rolling out age assurance in app store in response to U.S. laws, following similar moves by Apple; officially launches the Gemini for Home voice assistant in early access in the U.S. on Nest devices set to English.
• Apple releases the first preview version of Swift for Android, allowing developers to code Android apps in the Swift programming language.
• OpenAI releases gpt-oss-safeguard, its open-weight reasoning models for safety classification tasks in partnership with Discord, SafetyKit and Robust Open Online Safety Tools; comes as Apple releases Pico-Banana-400K, a comprehensive dataset of 400,000 curated images that's been specifically designed to improve how AI systems edit photos based on text prompts.
• NVIDIA makes history as the world's first company to reach US$ 5 trillion in market value, powered by a rally that has consolidated its position at the centre of the global artificial intelligence (AI) boom.
• Character.AI plans to phase out romantic AI chats for minors by November 25, 2025, and limit under-18s to two hours of conversations per day, a year after a 14-year-old committed suicide after forming sexual relationships with chatbots and amid increasing pressure to better safeguard younger users from potential harm.
• Meta's Threads hits 150 million daily active users, up from 100 million in December 2024, as the company explores algorithm personalisation controls and video ads for the platform; plans to add a "huge corpus" of AI content into its recommendation system to see even more AI generated posts on Facebook and Instagram.
• Grammarly rebrands as Superhuman after its acquisitions of Superhuman Mail in June and Coda in December 2024; releases an AI assistant called Superhuman Go that's built into Grammarly's existing extension to provide writing suggestions, give feedback on emails, and connect with other apps like Jira, Gmail, Google Drive and Google Calendar.
• European company Bending Spoons agrees to buy AOL from Yahoo; marks its latest purchase in a string of other brands like Komodo, WeTransfer, Vimeo, Evernote and Meetup.
• Vibe coding platform Cursor releases first in-house LLM, Composer, to execute coding tasks quickly and accurately in production-scale environments.
• Google updates Play Store policy in the U.S. to let developers link to external app stores and payment options in the aftermath of an court order in its long-running legal battle with Epic Games.
• F-Droid once again hits back at Google for enforcing new sideloading requirements, stating the "developer verification decree effectively ends the ability for individuals to choose what software they run on the devices they own." (The development comes amid a broader "Keep Android Open" movement that has pushed back against Google's proposed changes.)
• Google says AI Overviews in Search drive meaningful query growth and AI Mode has 75 million daily active users; adds live audio, instant transcripts and AI insights to Google Finance as part of a new earnings experience.
• Meta-owned WhatsApp announces a new way to access encrypted backups with passkey support for Android and iOS; gets time until the end of the year to allow users in the Netherlands to set a chronological, non-algorithmic feed as the default on Instagram and Facebook following case a brought by digital rights group Bits of Freedom, which argued that the company's current design violates the European Digital Services Act (DSA).
• Google partners with Reliance to offer free AI Pro access to Jio 5G users in India for 18 months. (It's worth noting that Perplexity has a similar deal with Bharti Airtel.); extends partnership with Magic Leap for another three years to develop Android XR glasses, while showing off a new prototype concept that combines Google's Raxium microLED light engine with Magic Leap's AR optics.
• Design platform Figma acquires AI-powered image and video generation company Weavy.
• Samsung releases a desktop version of its web browser for Windows, as it describes it as "evolving from a PC browser that waits for input to an integrated AI platform"; says it's partnering with Nvidia to build an "AI Megafactory" and deploy over 50K of Nvidia's most advanced GPUs to embed AI in its chipmaking process.
• GitHub says it has more than 180 million users, hosts more than 630 million projects, and that TypeScript has now become to top programming language on the platform, dethroning Python.
• YouTube revamps its TV app to mimic paid streamers like Netflix, organising videos by seasons and episodes, blurring the lines between traditional TV and creator shows.
• Perplexity launches a free AI-powered patent research tool in beta to let users search for patents using natural language queries; signs a multi-year licensing agreement with Getty to display its images across its search and discovery tools.
• X rival Bluesky hits 40 million users; introduces a new "dislikes option in beta to improve personalisation.
• China enacts new law as of October 25, 2025, that requires influencers in the country to show real-world qualifications before posting about sensitive topics like medicine, law, education and finance, or risk facing fines.
• Microsoft previews a new Windows 11 feature that allows shared audio, making it possible to stream audio across two pairs of wireless headphones, speakers, earbuds or hearing aids.
• Google says it has removed its AI model Gemma from AI Studio after U.S. senator Marsha Blackburn accused it of fabricating accusations of sexual misconduct against her; starts rolling out a new version of its Translate app with a feature that allows users to choose between faster or more accurate Gemini AI-assisted translations.
• The U.K. CMA says it will launch an in-depth investigation into Getty and Shutterstock's proposed merger after their proposed remedies failed to address concerns.
• Apple releases 26.1 of its software with an new "tinted" option for the Liquid Glass design to reduce transparency and increase contrast; refreshes the web version of the App Store, letting users view and search for apps for the iPhone, iPad, Mac, TV, Watch and Vision Pro.
• Apple unveils new colourful logo for its Apple One subscription bundle, shortly after it rebrands Apple TV+ to Apple TV.
• Amazon announces a new multi-year, US$ 38 billion cloud partnership with OpenAI, as part of which Amazon Web Services will provide the company with access to thousands of NVIDIA GB200 and GB300 GPUs for inference and training its next-generation models.
• A new investigation from the Columbia Journalism Review finds that OpenAI's ChatGPT Atlas browser takes great pains to avoid certain sources of information, specifically that belong to companies suing it for copyright infringement; says the application went out of its way to summarise articles from PCMag and the New York Times by looking up citations of the article from social media and other news sites.
• Apple updates its Podcasts app to include automatically-generated chapters for shows in English and allow creators to add links at specific timestamps in their episodes.
• Amazon sends a cease-and-desist letter to Perplexity accusing it of computer fraud and demanding that the AI startup stop its Comet browser from making purchases on the e-commerce platform on users' behalf; Perplexity calls the legal threat "aggressive" and a "threat to all internet users," adding Amazon is resorting to bullying tactics to "block innovation and make life worse for people."
• Google unveils Project Suncatcher to launch two solar-powered satellites, each with four TPUs, into low Earth orbit in 2027, as it seeks to scale AI compute.
• ByteDance-owned TikTok comes under investigation in France for failing to comply with requirements to vet material that promotes suicide on the Chinese video-sharing platform.
• Google and Epic Games reach an agreement to settle Epic's 2020 antitrust lawsuit, with proposed changes to Android and Google Play pending court approval; includes reduced app store fees to 20% or 9% (depending on the transaction) globally and a new program for alternative app stores to become first-class Android citizens through a program called Registered App Stores.
• Australia's eSafety Commission expands the country's kids social media ban to Reddit and video streaming service Kick, the others being Facebook, Instagram, Snapchat, Threads, TikTok, X and YouTube. (The new law, which goes into effect on December 10, 2025, requires companies to block Australian users under the age of 16 from their platforms.)
• ByteDance pulls the plug on Anthropic's Claude artificial intelligence models on its coding app Trae after the U.S. technology firm begins restricting services to Chinese-owned entities anywhere in the world.
• Apple plans to update iOS to allow users in Japan to install alternative app marketplaces on their devices; to also add a prompt to select a preferred search engine.
• A new report from Reuters finds that Meta is making billions of dollars every year from ads marketing scams and illegal products on its platform; internal Meta documents estimate that scam ads could account for as much as 10% of its revenue, a total that would amount to about US$ 16 billion.
• Meta allowed "high value accounts" to "accrue more than 500 strikes without Meta shutting them down," Reuters reported. The more strikes a bad actor accrued, the more Meta could charge to run ads, as Meta's documents showed the company "penalised" scammers by charging higher ad rates. In response, Meta said the estimate was rough and overly-inclusive, and that it has removed more than 134 million pieces of scam ad content so far in 2025.
• TikTok says fraudulent sellers on TikTok Shop are using AI to create fake brands and non-existent products; claims it rejected 70 million products and banned 700,000 sellers in H1 2025.
• Google rolls out a dedicated form to allow businesses listed on Google Maps to report threat actors who post bad reviews and demand ransoms to remove the negative comments.
• Italy becomes the latest to join a growing list of countries banning or restricting access to pornographic websites; introduces a new age verification rule for pornographic websites accessible within the country starting November 12, 2025, or face fines of up to €250,000.
• OpenAI faces seven lawsuits in the U.S. state of California, including four wrongful death lawsuits, claiming ChatGPT encouraged dangerous discussions, leading to suicides and harmful delusions; gets accused of rushing its product to market without adequate safeguards and acting as a "suicide coach" instead of people toward professional help.
• "Our safeguards work more reliably in common, short exchanges," OpenAI noted in a post in August 2025. "We have learned over time that these safeguards can sometimes be less reliable in long interactions: as the back-and-forth grows, parts of the model's safety training may degrade."
• OpenAI says the Cameo feature in Sora could allow companies and marketers could eventually charge for use of their "beloved characters" in its Sora 2 text-to-video app as a way to tackle copyright violations.
• Non-profit Common Crawl, which scrapes billions of web pages to build a massive archive of the internet, gets accused to quietly funnelling paywalled articles to AI developers. (According to Originality.ai, the Common Crawl's CCBot has become the scraper most widely blocked by the top 1,000 websites. The non-profit has claimed that publishers are making a mistake by excluding themselves from "Search 2.0.")