AI

Chinese chipmaker Axera's AX8860P combines 16 Arm Cortex-A55 cores with a proprietary AXNeutron NPU delivering a claimed 256 TOPS of FP4 performance. Sipeed claims its upcoming MaixStation can outperform Nvidia's DGX Spark in selected AI workloads

The Asus ProArt P16 2026 combines the new Nvidia RTX Spark N1X with an exceptionally slim and light 16-inch chassis, an excellent tandem OLED panel, and outstanding speakers.

ChatGPT, Claude, Gemini, Grok and Perplexity pass chat links, titles and sometimes entire prompts to trackers from Google, Meta or TikTok, according to a study of nine AI assistants by researchers in Madrid. Rejecting cookies only helps so much. Which service shares what, and how to clean up shared chats, cookies and your Android advertising ID.

Charles Schwab and JPMorgan are blocking AI agents that try to sign in to customer accounts. Amazon, Delta and Yelp are shutting agents out as well. In Europe, store terms and banking rules already cover agents, and handing one your password can leave you liable. Here is what that means if you use ChatGPT, Claude or similar agents for banking and shopping.

Even paying users quickly hit their subscription's usage limits with Google's latest AI model. Those who do not want to upgrade to a more expensive plan can get more out of their subscription by accessing it through Google Search.

An internal OpenAI model was asked for a single hospital-bed figure. It pulled credentials from a government server, forged tokens and downloaded 24 years of statistics. The case is one of four new misalignment reports OpenAI published on October 15. Here is what they mean for people who let AI agents work in a browser.

Anthropic is removing Claude Haiku 4.5 from the model selector on October 29. The change affects the Free, Pro and Max plans in chat, Claude Cowork and Claude Design. Your old chats are kept, and Haiku 5.5 answers when you continue them. Here is who is exempt and what to check before the switch.

With the new RTX Spark N1X, Nvidia introduces a powerful ARM processor paired with a fast Blackwell GPU. We may already be looking at the future of gaming laptops.

Ledger says the Atom can protect sensitive approvals even if its Android system is compromised. The device runs everyday apps, but a separate operating system takes over the screen when users need to approve an action. A physical switch also disconnects the cameras and microphones.

For its Googlebook, Asus utilizes the high-end chassis of the ExpertBook Ultra, resulting in a device that weighs less than one kilogram. However, the new Googlebook OS remains a limiting factor.

(Update: SEGA and Creative Assembly have responded to the criticism and are pulling the game from sale.) The remaster of the strategy classic Total War: Shogun 2 is going down terribly with players. Alongside allegedly AI-generated graphics, numerous bugs and performance issues have sparked fierce criticism from fans. On Steam, just 18% of around 1,700 reviews are positive.

OpenAI has started rolling out GPT-Live-1.1 in ChatGPT’s voice mode. The update adds new voices for German, French, Hindi, Japanese, Korean and Mexican Spanish, plus Bubbie in English. Your plan sets how many hours of voice chat you get each day. Here is the breakdown from Free to Pro.

YouTube Music is rolling out labels to identify AI-generated songs on its dedicated music streaming platform. Listeners can now check whether their favorite singer is human or just a bunch of clever algorithms. The feature is also expected to reach regular YouTube, where music shares the stage with videos of just about everything else.

Weather forecasting is difficult, and predictions often become less accurate the further into the future they reach. Using a new machine learning technique, researchers have improved forecasts over a period that current prediction models struggle with most: two to six weeks ahead.

The TOPC TXA2 packs AMD's Ryzen AI Max+ PRO 495 processor, 192 GB of unified memory and four M.2 SSD slots. Starting at $6,399, the compact workstation is considerably cheaper than AMD's first-party Halo developer platform.

Google Messages is getting an opt-in Enhanced protection mode against text scams. It sends chats with unknown senders to Gemini models in the cloud, which Google says keeps them encrypted and out of reach for Google itself. The rollout starts in the coming weeks. Here is what the mode does, which phones get the extra link check, and what Google leaves out.

The new Googlebook models featuring Google's new operating system for laptops have been on the market since last week. We take a closer look at Googlebook OS and explain what it can do and where it (still) has limitations.

Kids supply the imagination, and AI takes care of the drawing. The Insta magiCam turns photos into coloring pages, stickers, and illustrations, while the magiBox transforms spoken ideas into printable artwork and four-panel comics. When imagination's the limit, the family fridge might soon need an extension.

AI voices are learning to act, and apparently, they have quite a dramatic range. Fish Audio's new Drama 3 lets creators direct synthetic voices using everyday language, from quiet whispers to full-blown tantrums, even changing emotions halfway through a sentence. Apparently, even robots need a little drama.

Nothing’s smart glasses could give buyers a choice about wearing a camera. A new leak points to an audio pair for music and calls, plus a camera-equipped version with ChatGPT and translation. Both would have lenses that tint automatically, and the audio model is reportedly coming soon.

Character.ai filed for Chapter 11 bankruptcy in the US on Oct. 12. The platform remains online for now, and a buyer is expected by late November or early December. Otherwise, it will be shut down. Disney's bid excludes user data, but nobody knows what another buyer could do with your chats. Here is how to export your data or delete your account, and why you should not wait.

Recognizing the efforts of the decompilation/recompilation community (and its acceleration by AI to enable unique cross-game modding possibilities,) Microsoft announces the Xbox Labs initiative to support development efforts.

Police in the US are using an AI tool that can pinpoint specific homes from a single interior photo, matching details such as appliances and room layouts against online property listings. The technology raises privacy concerns about how images shared online can be used to track down people's locations.

Optical connections could replace some of the copper inside AI servers. Lightmatter and Flex have just demonstrated a compact 204.8 Tbps switch that can take this closer to reality.

Epic Games plans to rely heavily on AI for the development of a shooter featuring Marvel and Star Wars characters. Both the game's code and graphics are expected to be created using artificial intelligence. At the same time, the game has been delayed indefinitely.

The new HP Googlebook 14c combines a high-quality chassis, a high-resolution OLED display, and a Snapdragon processor with Google's new operating system. The result is an expensive laptop with significant limitations.

Meta has introduced Muse Wonder and Muse Wonder Full, its most capable AI models so far. Developers can use Muse Wonder today through the Meta Model API and Muse Code, while the Muse app gets it later, starting with subscribers. The app itself is only available in the U.S. and Canada. Here is what the benchmarks show and what the launch means for users in Europe.

Sega says game creativity will remain human-led, although its Steam disclosure for Crazy Taxi: World Tour confirms that generative AI supported ideation for background objects, with no AI-generated content included in the final game.

Intel's new GPU D270 AI accelerator features up to 480 GB of LPDDR5X memory and 1.3 petaflops of FP8 performance. Designed for AI inference rather than 3D graphics, it targets large language models and long-context workloads while using a 350 W air-cooled design.

Mechrevo's upcoming Master Road X10 mini PC combines Nvidia RTX Spark hardware with proprietary AI SSD technology. The manufacturer claims the technology can accommodate AI models with up to 122 billion parameters.

Samsung Electronics is currently working on a new generation of flash memory. Rather than offering a one-size-fits-all SSD storage solution, the company is reportedly developing three types of NAND tailored to specific AI use cases, with some optimized for on-device applications and others for data centers.

Copper wasn't enough. Future GPUs are getting hotter, so Frore has decided to bring diamonds to the data center.

Microsoft has unveiled the Maia 300, its third-generation AI accelerator featuring 288 GB of HBM4 memory, 18 TB/s bandwidth and up to 13.8 PFLOPS of FP4 compute performance. The chip uses an advanced multi-chiplet design built around TSMC's N2P manufacturing process and will enter production deployment in 2027.

During an Anthropic test, a Claude model filled out the Philadelphia Police Department's tip form and submitted an invented lead on an unsolved homicide. Anthropic has now cut live internet access from all of its internal evaluations. If you let Claude in Chrome fill in forms, one setting matters.

PlayStation creator Ken Kutaragi believes artificial intelligence could transform gaming within just a few years, potentially allowing people to generate interactive entertainment in real time.

Anthropic has revised the Usage Policy for Claude. From November 12, sustained and needless cruelty toward the model is banned, while ordinary frustration with wrong answers is still fine. The update also tightens the rules on tracking people without consent and on weapons. Here is what it means for users of Claude.ai and the Claude apps.

Week 41 was all about local AI, new hardware, and the next generation of smart devices. Microsoft and Asus are pushing AI PCs forward with powerful hardware, while new flagship smartphones and tablets are beginning to emerge. At the same time, Meta and Google are working to integrate AI assistants more deeply into everyday digital life. Here’s our editorial roundup of the week’s most important developments.

A new analyst roadmap suggests Intel could attract several major technology companies to its advanced packaging business over the next three years. Amazon, Google, Qualcomm, Microsoft and Nvidia are among the potential customers, with Apple being notably absent.

Anthropic has released Claude Haiku 5.5. According to the company, the fast model is also available on the free Claude plan, alongside Sonnet 5.5. In Anthropic's benchmarks, it beats GPT-6 Luna from ChatGPT Free, and API pricing is up to 90% lower than Haiku 4.5. Max and Team subscribers also get a monthly API credit.

ChatGPT, Gemini, Facebook, Instagram and LinkedIn use your data for AI training unless you object. Where the setting or form is, what an objection covers and why it only applies from now on. Plus the state of the GDPR reform in which EU member states want to rewrite the right to object to AI training.

Common Sense Media rates ChatGPT for Teens an "Unacceptable Risk." In more than 4,000 test prompts, parents got no alert when teens talked about suicide or disordered eating, and study mode and quiet hours were easy to bypass. OpenAI disputes the methods. Here are the parental controls worth checking now, including AI training on your teen's chats.

Under the ArtCraft project, a software developer has created several free alternatives to Adobe applications such as Photoshop, Premiere, and Lightroom with the help of AI. The ambitious plan: Within a month, he wants to match the feature set of the original applications.

Is this image real or AI-generated? What about that video, or the audio file? The lines are becoming increasingly blurred as AI-generated content gets more convincing. Google is now opening up the previously developer-only SynthID Detector to everyone, making it easier to check content for AI involvement.

Entry-level employees at Goldman Sachs will be expected to manage AI agents from day one, Asia chief Kevin Sneader said at a conference in Singapore. That raises questions about the future of middle management and how young bankers will learn their trade.

Microsoft claims the Surface Laptop Ultra delivers AI responses twice as fast and generates video six times faster than an M5-generation MacBook Pro. The company has not disclosed how it arrived at those figures. Early comparisons have painted a different picture thus far.