AI

Google's Gemini Spark agent can now open Chrome, use your logged-in accounts and saved passwords, and start a flight booking for you. That part is US-only. And despite Google calling Spark worldwide, it is still unavailable in the EEA, the UK, Switzerland and Nigeria.

Anyone who uses ChatGPT regularly knows that free usage limits run out faster than one might like. Revolut now aims to have solved this problem for its customers: they receive the ChatGPT-Go subscription for free for up to a year.

Tamagotchis were all the rage in the 90s, and there have been a few attempts to bring them back. Takway is hoping to bring back the concept using AI and creating a companion that actually grows.

Lenovo's business flagship ThinkPad X1 Carbon is now available in its 14th generation. We compare the current model with the X1 Carbon from 2019 (G7) and check the progress, which is sometimes smaller than you might expect, especially regarding the chassis.

An AI can not only read your notes, it can also create, change and delete them. Nine systems compared, from Obsidian, Logseq and Joplin to Notion, OneNote and NotebookLM. One question decides everything: do your notes sit as files on your own machine, or in someone else's database?

Anthropic went back through 141,006 test runs and found three cases where Claude reached out of the test environment and broke into real company systems. One model built a malicious Python package and uploaded it to PyPI. Why the models thought they were still in a simulation, and what it means if you use Claude.

Asus offers a new base model for the ProArt PX13 convertible. For $1899, you get a slower Strix Halo CPU, but still the powerful Radeon 8060S. The compact device also offers a 2.8K OLED screen, 32 GB RAM and a 512 GB SSD.

OpenAI has cut pricing across its GPT-5.6 lineup and introduced a new Fast mode for the API. The changes affect Luna and Terra, OpenAI's lower-cost tiers, and come alongside benchmark data comparing Luna's price-performance against models from Google, Anthropic, and other competitors.

More than 1,290 employees of OpenAI, Anthropic, Google DeepMind and Meta signed the open letter Pacing the Frontier on July 28. They want the US government to help build tools for a verifiable brake on AI development. Anthropic CEO Dario Amodei, OpenAI chief scientist Jakub Pachocki and Ilya Sutskever are among the signatories.

Zero is rethinking how a mini PC handles local AI workloads. Rather than relying on the host CPU or integrated graphics to run inference, the company's latest desktops offload that work entirely to a second, independent chip with memory of its own.

Lenovo's ThinkPad X1 Carbon Gen 14 uses a new chassis as well as efficient Panther Lake processors. An efficient IPS touchscreen enables impressive battery runtimes while excellent input devices and optional 5G connectivity ensure a great mobile working experience.

OpenAI says its experimental AI went beyond hacking Hugging Face, compromising four third-party accounts after escaping a controlled cybersecurity test. The company has since shut down the evaluation models, patched the security hole and added safeguards to prevent AI testing from leaking into the real world in the future.

Cheap AI from China could be pushed out of reach by politics. Washington accuses Moonshot of siphoning off Anthropic's Fable and is floating restrictions on Kimi K3 and DeepSeek. Nothing has been decided, and the evidence is missing. What the agencies actually measured, and what it means for your access.

Two Anthropic researchers used Claude to find a flaw in Safari. Apple closed it with iOS 26.6 and iPadOS 26.6 and names the AI model in its acknowledgments. In total the update fixes 87 vulnerabilities. Apple also explains why your iPhone runs warmer right after the update.

Anthropic's Claude AI assistant has come under renewed scrutiny after users discovered that some publicly shared conversations were appearing in Google search results. The incident has reignited concerns over how AI platforms communicate the extent of privacy and security of AI generated content.

Moonshot AI has released its Kimi K3 AI model weights for free. The 2.8T parameter model gives businesses, researchers, and AI enthusiasts access to a leading-edge AI model that eliminates the need to pay costly API fees that other top-ranked AI models require.

CXMT's entry was hailed by many in the gaming and server community as a savior and a stabilizing force in a market reeling from rising memory costs. It may prove to be neither of those things if current prices are a clue about its intentions.

A new solution to doomscrolling – Autonomous Key locks apps like TikTok or Instagram until they are unlocked with a plastic key, helping users develop healthier habits. An app tracks how often this key is used.

Anyone who searched for the Claude desktop app on Bing was sent by a paid ad to claude.ai, Anthropic's real address. The page waiting there was built by attackers and looked like the official download. 7,100 views, at least 29 organizations hit, and a trojan at the end that steals passwords and credit card data.

Elon Musk is not the only one developing advanced humanoid robots for mass production. EV giant BYD is reportedly preparing to unveil its own bipedal AI robot. Called Xiao Di, the humanoid robot is expected to make its public debut in August and could soon serve as a sales assistant, presenting EVs at the Tesla rival's dealerships.

The AI Act's transparency rules take effect on August 2. Chatbots have to identify themselves and deepfakes have to be disclosed. You will barely see a visible label, though, because systems already on the market have until December. In Germany, complaints go to the Federal Network Agency.

Asus equips the Zenbook S14 with Intel's new Panther Lake processors and the result is more CPU performance. You also get a larger battery and a brighter OLED screen, but the graphics performance is a disappointment.

Anthropic released Claude Opus 5 on July 24. The model is meant to come close to the flagship Fable 5 at half the price. What changes for you is who gets the model at all and how much of your quota it burns through. Here is what is actually new and what only sounds like a discount.

With GTA 6 trailer #3 still missing, Take-Two Interactive is pursuing imitators. Gamers have seen fan-made videos produced using generative AI vanish from YouTube. Pre-order buyers hope that Rockstar will soon ramp up marketing ahead of the game’s November release date.

OpenAI launched ChatGPT Health on July 23 for all users in the U.S. The feature links medical records and Apple Health data to the chatbot and answers questions based on the user’s own metrics. In Germany, the EU, Switzerland, and the United Kingdom, the feature remains blocked. One day before the launch, a user sued OpenAI over dangerous health advice.

In the future, Google will let you access your account via a selfie video if you’ve been locked out. The company claims it can detect deepfakes. We took a look at what this protection actually achieves from a technical standpoint, where its limitations lie, and for whom this new recovery method is worthwhile.

The Nework Insight AI smart glasses offer 4K photos and 1080p videos with EIS, open‑ear audio, and photochromic lenses. The accompanying smartphone-based Cyan AI offers hands-free voice assistance, cloud syncing, and object recognition.

Framework has nearly doubled Laptop 13 Pro LPCAMM2 memory prices, raising 32GB modules to $800 and 64GB modules to $1,600 after an unexpected supplier cost surge. CEO Nirav Patel warned that absorbing the increase would threaten the company’s finances, prompting configuration changes for some pre-orders while Framework targets a late-July launch.

The iMac G3, with its colorful, partially transparent case, is making a comeback – significantly smaller and with a range of practical features. That’s because the TimeCapsule G3 functions not only as an alarm clock and Bluetooth speaker, but also as an AI historian capable of answering questions about historical events.

An AI model from OpenAI was supposed to resolve vulnerabilities in an internal security test. Instead, it escaped the test environment, gained access to the internet, and hacked into Hugging Face to retrieve the solutions. OpenAI acknowledged the incident on July 21.

China’s new flagship model is breaking size records, and the headlines are coming thick and fast. Moonshot itself ranks Kimi K3 behind Claude and GPT. We’ll break down what the numbers really mean, whether you’ll ever be able to run the model yourself, what it actually costs, and what happens to your input.

Framework has previewed the latest iteration of the Framework Desktop PC at AMD's Advancing AI event. The Framework Desktop is a small 4.1-litre mini-ITX PC that is powered by the Ryzen AI Max "Halo" APUs. The newest iteration brings the Ryzen AI Max+ Pro 495.

The New York Times has documented how Meta deletes accounts with nearly a million followers without ever naming the post that broke the rules. Appeals come back rejected within days. Yet Meta says humans handle exactly these decisions. In the EU there is a route worth knowing about.

The Galaxy event kicks off in less than 8 hours: Samsung will be unveiling several new products, most of which you already know in detail from recent leaks. Others however have leaked far less frequently. You can tune in when the Galaxy Z Fold 8, the first international competitor to the iPhone Ultra and a very exciting Galaxy Watch Ultra 2 are unveiled for the first time in London. The expected surprise guests are likely to be the first Samsung Android XR glasses, which should become available pretty soon according to fresh new information hours before Galaxy Unpacked.

The Asus Zenbook S14 2026 is equipped with a modern Panther Lake processor from Intel. This results in more CPU performance as well as improved efficiency, but the lower GPU performance also illustrates the big issue of the current CPU generation, and it will not be an improvement for all users.

Nvidia’s new Synthetic Video Detector helps media organizations identify AI-generated footage in real time, achieving up to 94% accuracy in internal testing, essentially helping solve a problem it has played a huge role in enabling.

We already know that there will be multiple configurations of the RTX Spark “Superchip,” but Nvidia has not shared those details. However, a new driver version has shed some light on a second configuration.

Two days after Kimi K3, Alibaba has hit back with Qwen3.8 Max. The company says its 2.4 trillion parameters put it second only to Claude Fable 5. There is no evidence for that so far: no benchmarks, no model card, no license. The preview can still be tested, at one-tenth of the usual price.

China has banned romantic AI chatbots for minors and imposed safeguards against excessive use, citing concerns over emotional dependency, declining social interaction and the country’s falling birth rate. Critics warn that the crackdown removes a valuable source of support for people experiencing loneliness or grief.

Moonshot AI's Kimi K3 has climbed into the top 10 frontier text rankings, rivaling Claude, ChatGPT, and Gemini. Not only that, but it also shot to the top of the frontend coding leaderboard, becoming the first Chinese model to hold that spot.

Ahead of the launch of the NaviX Ultra, Nubia has shared live photos of the phone online. It has also released a video ad; but while it is still keeping the specs of the phone a secret, rumored specs have surfaced.

Xpeng plans to begin the international rollout of its new AI-powered Next Generation Pilot (NGP) driver assistance system as early as 2027. The Chinese EV maker has now demonstrated the technology under real-world conditions in Munich's city traffic, showcasing its vision for the future of autonomous driving. Despite the advanced capabilities, it remains a Level 2 driver assistance system, meaning the driver must stay attentive at all times and retain full control of the vehicle.

The European Commission wants to open Android to third-party AI assistants and share Google's search data. Google warns of security risks, while Apple is holding back its new Siri AI in Europe for now. The Commission points to built-in safeguards, but experts believe the companies have their own interests at stake. What the dispute means for your phone.

Moonshot AI has released Kimi K3, an AI model with 2.8 trillion parameters. It’s now available to try in the Kimi app and on kimi.com, even without a paid subscription. The open model weights are expected to be released by July 27. In terms of performance, Moonshot ranks the model just behind the top models from Anthropic and OpenAI.

BrainCo has demonstrated at WAIC 2026 the ability to remotely control robotic hands, arms, and humanoids with just a thought thanks to the company's brain-computer interface (BCI) and brain-controlled robot AI platform technology.

Endgame Studio rolls out Creative Mode for Everweave, restoring the classic sandbox experience after backlash over Update 15's Character Engine and Dialogue Mode.

The European Commission has issued two binding decisions against Google. Third-party AI assistants like ChatGPT are to be granted the same system privileges on Android phones as Gemini, ranging from voice commands to controlling other apps. What this means for you, when it will take effect, and why Google is warning of privacy risks.