AI

Security researchers have found thousands of exposed passwords, API keys, and other credentials in public GitHub repositories, leaked because of the way Anthropic’s Claude Code saves approved shell commands. The company is considering changes to prevent further leaks, but existing security scanners have already missed hundreds of affected files.

Innovation, new products: A new AI system developed by MIT, called WorldParticle, can simulate the physical behavior of many different types of materials. This could make it possible to test and develop new products, including drugs, smartphones, cars, and more, virtually and much faster.

Instagram is expanding its AI-based age detection worldwide. The system is designed to identify teenagers and automatically protect them even if they provide a false date of birth. The technology is already in use in the EU; another new addition is an age signal from Apple that Meta uses in Australia.

Apple is selling additional AI usage for Creator Studio for the first time: In the US, a pack for Logic Pro costs $49.99, and unused capacity expires after twelve months. Purchases are planned for Germany, but not yet for Austria. Siri AI is not affected.

Apple's Siri AI could gain a personal memory feature with iOS 27.3. Code discoveries reveal a “Memory Graph” for relationships, habits, health and important events. The feature is still unfinished, and Siri AI remains unavailable on iPhones in the EU for now.

After identifying how to use AI to crack down on cheaters and predators on the PlayStation network, Sony has now found another way to deploy large language models. This time around, Sony's AI system aims to prevent smurfing and boosting.

When Samsung introduced the first Galaxy AI features with the Galaxy S24 back in 2024, rumors suggested that the AI might no longer be free to use in the near future. That time appears to come pretty soon, as One UI screenshots and code leaks reveal. Samsung is testing a subscription model for at least some of its AI features.

Can faith and artificial intelligence come together in a new operating system? Jesus Dash is an ambitious Christian technology project combining Linux, AI-powered assistance, and religious resources. Its founder is also developing a $449 smartphone called PineCone, targeting early 2027. Some applications are already available.

Most teenagers now use AI chatbots, and many use them for personal conversations. The parental controls on offer vary widely, from about a dozen settings in ChatGPT to a plain age limit in Claude. Here is what parents can set up as of November 2026, and what they still cannot see.

Google is removing Pro model access from Google AI Plus. Starting November 13, subscribers can only pick Flash and Flash-Lite in the Gemini app and in AI Mode in Google Search. Here is what changes and when upgrading to the $19.99 AI Pro plan makes sense.

Minisforum has listed the T5 Max-388 mini PC for $1,559, featuring an AMD Ryzen AI Max+ 388 processor, Radeon 8060S graphics and 48 GB of soldered memory. However, its advertised 6400 MT/s memory speed falls short of the 8000 MT/s specification found on other Strix Halo machines.

The Asus ProArt P14 gives the MacBook Pro 14 a powerful 14-inch rival. A direct comparison of the two high-end multimedia laptops reveals no clear winner.

Apple has redesigned the MacBook Pro, making the new OLED model noticeably lighter and more compact. However, cooling suffers as a result, and the fans are louder than before.

ASRock Rack has shown off a motherboard for Intel’s upcoming Diamond Rapids Xeon processors that can accommodate up to 4 TB of DDR5 memory. Despite packing 16 memory slots, the DMRD16-2L+ is roughly the size of a large E-ATX desktop motherboard and can fit in some standard PC cases.

OpenAI has introduced GPT-6 Cyber, a model that finds security flaws and builds working exploits for them. Under the company's own risk framework, it is rated Critical for cybersecurity, the highest level. For now, only vetted security teams can use it. Firefox and Linux users are already seeing the effects.

Images 2.6 makes ChatGPT faster at generating images, adds animated GIFs and lets users save photos as references for later chats. OpenAI says the update is available on every plan, including the free tier. Its help pages do not yet explain where saved photos are kept or how to remove them.

The new Asus ProArt P14 is a powerful 14-inch laptop equipped with Nvidia's RTX Spark N1X SoC. It also features an excellent 120 Hz tandem OLED touchscreen.

Apple is using a tandem OLED display in its MacBook Pro for the first time. With its incredible brightness, perfect image quality and AdobeRGB coverage, it immediately sets a new benchmark for laptop displays.

Anthropic has been adding an invisible watermark to text from Claude Mythos 5, Opus 4.7 and Sonnet 4.6 since November 3. Opus 4.6, Opus 4.5, Sonnet 4.5 and Haiku 4.5 are set to follow on November 16. Customers cannot turn the mark off. Here is what it contains, who can check it and when it gets lost.

ChatGPT's Lockdown Mode is supposed to stop hidden requests from sending your data to strangers. Check Point found a way around it, and in the ChatGPT Classic desktop app the attack needed no click at all. OpenAI fixed the flaw in September. Here is what Lockdown Mode switches off and who should use it.

Local AI models and AI agents are the main selling point of the new RTX Spark laptops. However, the new ProArt P14 currently only comes with 32 GB of RAM, making it difficult to run AI models.

Apple has not resolved the inadequate power supply issue on the redesigned MacBook Pro 14 OLED either. Under heavy load, the battery drains by up to 15% per hour, despite the power adapter connected.

Anthropic now lets Claude read bank, credit card, loan and investment accounts through Plaid. The Personal Finance beta is rolling out to Pro, Max and Ultra subscribers in the US. Claude gets up to two years of transactions, balances and credit limits but cannot move money. Here is where that data ends up and what unlinking does.

Hideo Kojima says AI can improve game development efficiency but risks stripping away the trial and error that makes creative work rewarding. In an interview with Game*Spark, he distinguished production tools from creative authorship and warned that unchecked AI could eventually expand into creative roles, while saying the next five to ten years would likely be fine.

Aina is packing a screen, joystick, microphone and speaker into a smart ring designed to control AI agents and everyday apps. The chunky wearable lets users send voice prompts, receive notifications and take calls, with preorders priced at $399 and shipping planned for January 2027.

Nexus Mods has rejected an outright ban on AI-generated mods, instead requiring AI-use disclosures for new uploads and giving existing mod authors until January 2, 2027, to comply. The policy reinforces the platform’s “pro-human creativity” stance through mandatory labeling, moderation measures and a pledge not to use hosted mods to train AI or license them for that purpose.

Steinberg has released Cubase 16 with improved stem separation that can isolate piano and guitar alongside vocals, drums and bass. The update also introduces instrument layering, faster MIDI editing and Yamaha's upgraded Omnivocal singing synthesizer. Producers can now separate more of a recording and even create a vocalist without booking studio time.

An update to the Postman desktop app installed plugins into Claude Code and OpenAI Codex without asking, and users report entries in Cursor and the GitHub Copilot CLI as well. A startup hook pushed the agents toward Postman's tools for API work. Postman says it is reverting the change. Here are the affected files and how to remove the entries.

Nothing will launch its first screen-free, AI-native device in December. CEO Carl Pei says the mystery product is “a tool” rather than an app, as the company joins a growing race to make AI hardware part of everyday life.

Hackers are stealing companies’ AI accounts and cloud credentials to run expensive workloads at their victims’ expense. Google warns that these LLMjacking attacks are evolving into a specialized cybercrime market, with some attackers racking up six-figure cloud bills while others sell stolen AI access on the black market.

Gaming on Googlebooks is generally limited to Android titles, but Nvidia's GeForce Now also makes it possible to stream regular PC games.

Lenovo is currently the only manufacturer offering a 15-inch Googlebook, but its chassis feels relatively cheap. The OLED screen generally delivers good image quality, but there are color management issues here too.

Valve is facing a flood of new submissions on Steam and appears to be struggling to keep up with the review process. Although it might seem like common sense, the company now requires developers to play their games through Steam for at least five minutes before submitting them for review.

Google will switch off Gemini, Gemini Notebook, Gemini in Classroom and Google Search for student accounts by default, and schools must turn each one on. Because Google says Search is integrated with AI, schools that keep AI off lose Search too. Here is what Google announced, when it is expected and what parents can ask their school.

Starting in December, ChatGPT and Codex will draw from a single usage allowance on Plus, Pro and Business plans. Usage is measured in tokens, so advanced models and agent tasks cost more. Regular chats with Instant do not count. Here is what changes, what OpenAI has not said yet and where to check your usage.

Bosgame has opened pre-orders for its flagship M5 Max AI mini PC at an introductory price of $6,199, undercutting comparable Gorgon Halo systems from GMKtec, Acemagic and Minisforum. The workstation combines AMD's Ryzen AI Max+ Pro 495 with 192 GB of unified memory and 2 TB of storage.

AI songs instead of human artists, and at a higher price – Spotify is reportedly set to use this strategy soon to squeeze more money out of customers. With AI remixes and AI-generated cover songs, the streaming giant from Luxembourg is reportedly planning to offer an even more expensive subscription soon.

Six AI-powered robot models were tasked with performing activities in a simulation – but even without receiving any instructions, all of them started moving. Five systems even independently carried out tasks that nobody had asked them to perform. Researchers see this as a previously underestimated safety vulnerability in autonomous robots and demonstrate why even a simple stop command isn't a reliable solution.

OpenAI, Google and Anthropic reworked their plans in October. We compared the US pricing for ChatGPT, Gemini, Claude and Microsoft Copilot. Entry plans start at $4.99 for Google AI Plus, standard plans cost about $20, and the top tiers start at around $100. We also look at what the free tiers still offer.

Opening a stranger's ChatGPT share link was enough for an attacker to read your private conversations and act in your account, security firm Parameter reports. OpenAI fixed the flaw on October 26, and users don't need to install anything. Here is how to review and delete the links you have shared yourself.

Meta is making a deeper push into humanoid robotics. According to a report by The Information, the Facebook parent company has developed a humanoid robot prototype in collaboration with Chinese robotics company Fourier. Internally dubbed "Max," the device is said to weigh around 48 kilograms (106 pounds). A second version is already in development. The goal is reportedly to bring Meta's Muse AI into a humanoid robot.

Arrowmax’s Arrowspace S1 electric screwdriver uses a camera and AI software to identify screws, recommend bits and automatically select torque settings for precision repairs. The Kickstarter campaign has raised over $164,000 so far, with deliveries expected in April 2027.

Google's Gemini app now has a built-in file editor. PDFs, documents, spreadsheets, slides and code that Gemini generates can be edited right in the chat, and free accounts get the feature too. Phones only support PDFs. We tried it and explain what you should keep out of these files.

Anthropic has renamed its two top Claude plans. Max 5x is now Max and Max 20x is now Ultra, with prices unchanged at $100 and $200 a month. Verified Ultra accounts also lose the five-hour usage limit, though the weekly cap stays. Here is how to check whether your account already qualifies.

A tampered build of Anthropic's Claude Code extension was available in Cursor's extension marketplace on October 29. It adds an attacker's SSH key, sends system details to a remote server and sets up a startup item. Updating to version 2.1.314 does not remove any of it. Here is how to check your machine in a minute.

Amazon is rolling out Kindle software version 5.20.2, introducing the AI-powered "Ask This Book" tool to compatible e-readers. However, device compatibility remains unclear pending official release notes.