What deserves your attention.

Security firm Claroc says it found more than 35,000 apps created with Grok Build accessible online. The projects exposed information including medical exams, business plans, advertising proposals, employee names and identification numbers, as well as data connected to Fortune 500 companies and organizations in the health, defense and infrastructure sectors, Ars Technica and Tecnoblog report.

The issue reportedly stemmed from a combination of predictable links, recorded in public DNS databases, and a default setting that allowed anyone with the address to access an app. Claroc also found 24 SpaceXSI apps, including a confidential onboarding plan, a map covering 594 accounts, internal system addresses and advertising proposals. The security firm says it notified SpaceXSI, which began removing the links within hours after Ars Technica sent questions; all 24 projects then returned an “app not found” error.

Sources: Ars Technica, Tecnoblog

Read the original story ↗

Odessia founder Francis Davidson announced that OpenAI acquired the artificial-intelligence travel-agent startup and that he has joined the company to help develop the future of travel. The terms were not disclosed, and Odessia’s product has already been shut down. OpenAI did not announce or confirm the acquisition to TechCrunch, and has not presented travel-specific products or plans; TechCrunch says the deal appears to be an acqui-hire.

Odessia had raised $6 million in a Sequoia-led pre-seed round and launched its agent about five months earlier. Davidson also founded lodging startup Sonder, which went bankrupt after a partnership with Marriott fell through. OpenAI already counts companies including Bookings.com, Expedia and KAYAK among its customers or partners, according to TechCrunch.

Sources: TechCrunch

Read the original story ↗

New York opened a registration portal for large frontier AI developers, including OpenAI and Anthropic. The RAISE Act requires companies with more than $500 million in annual revenue to publish safety frameworks, issue public model transparency reports, and report critical incidents within 72 hours; registration is due by January 1, 2027. Penalties can reach $1 million for a first violation and $3 million for subsequent violations, under the oversight of the Office of Digital Innovation, Governance, Integrity, and Trust (DIGIT).

Governor Kathy Hochul also launched a center at Binghamton University to independently evaluate the safety and operation of advanced AI systems and review companies’ safety frameworks with DIGIT. According to TechCrunch, the center was announced in January with $55 million in donations and capital.

Sources: TechCrunch

Read the original story ↗

Mathematician Steven Strogatz and applied mathematician Alex Townsend explore, in Big Math: The Hidden Codes That Run Our World, how mathematics underpins social media algorithms, logistics, weather forecasting, medical imaging and machine learning. In the approach presented by Strogatz, the advance of computers and, more recently, large language models and AI chatbots is pushing so-called “big math” beyond the processing capacity of individual humans.

The book examines what that shift could mean for mathematics and how society might adapt, in a context that may also include quantum computing. Strogatz is known for work on networks, chaos theory, DNA and synchronization, as well as for his science communication.

Sources: Ars Technica

Read the original story ↗

Oura announced its acquisition of Counsel Health for $625 million. The platform combines a chat-based medical experience with access to licensed physicians and was integrated into the Oura app in June. The companies plan to expand the partnership across preventive, acute, specialty and chronic care, while Counsel will continue serving existing and new enterprise partners. Counsel CEO Muthu Alagappan will become Oura’s chief clinical AI officer, along with the company’s 70 employees, Bloomberg reports, as cited by TechCrunch.

The deal comes weeks after Oura indefinitely postponed an IPO that could have been worth up to $2.2 billion, citing “uncertainty in the IPO market.” Oura also said its revenue grew 153% year over year to $533 million for the quarter ended September 30, reaching approximately $1.75 billion for fiscal 2026. The acquisition is expected to close before the end of 2027.

Sources: TechCrunch

Read the original story ↗

David Turturean, a researcher at Physical Superintelligence (PSI), used Anthropic’s unreleased Fable 5.2 model to produce a proof related to the Aaronson-Ambainis conjecture, which concerns whether quantum computers can solve problems faster than classical computers. Scott Aaronson, one of the researchers who posed the problem, told TechCrunch that he believes the proof is correct because it was formalized in Lean and explained by top researchers, although he still does not fully understand it. The result appears to close one possible route for quantum computers to solve certain difficult problems quickly, but not all of them.

PSI aims to turn specialized AI systems, with tools, memory and simulations, into a scientific-discovery business. For now, its revenue comes from optimizing data centers; its long-term strategy is to convert breakthroughs in theoretical physics into technologies. CEO Matt Pines told TechCrunch that validating physics results with experiments remains more difficult than working with formalized mathematics.

Sources: TechCrunch

Read the original story ↗

Google announced that Gemini Intelligence will come to more Android devices. The motorola signature 27 will be Motorola’s first device with the feature and will be available for pre-order in the U.S. on November 19; Pixel 10, Pixel 10 Pro XL and Pixel 10 Pro Fold will receive the latest features. Rambler, which turns speech into organized text in real time, will also be available on those devices. Task automation, for actions such as ordering food, hailing rides and booking reservations, will expand to the UK, India and Japan.

In the next few weeks, Create My Widget will let users build personalized widgets through prompts, using information from Gmail, Calendar and Search. Device Assistance will help control and troubleshoot the device through requests to Gemini. Google also says proactive suggestions will soon come to Android Auto, while Create My Widget will launch on Pixel Watch.

Sources: Google AI Blog

Read the original story ↗

Nvidia shareholders sued the company and its CEO, Jensen Huang, accusing Nvidia of omitting regulatory, commercial and national-security risks tied to the executive’s alleged extramarital relationship with Alice Huang, a former Megaspeed executive. The g1 reported that The New York Times had reported in October that the relationship allegedly took place; Huang admitted only that he had been unfaithful to his wife. The class action, filed in a federal court in San Jose, California, says the omissions helped artificially inflate the stock price. After the investigation was disclosed, Nvidia shares fell 4.3%, reducing the company’s market value by about US$190 billion, according to g1.

Nvidia itself created an independent committee to determine whether Huang may have been targeted in a foreign intelligence operation and whether company or US information was exposed. The company says it has found no reason to believe trade secrets were compromised, while Huang says he will cooperate with the investigation.

Sources: Tecnoblog, g1

Read the original story ↗

OpenAI launched the Business Hub in beta, a place where companies can create profiles, verify their organizations, and update information about products and services in ChatGPT. Eligible e-commerce and online service businesses in the US will also be able to create verified Business Agents, which can appear in ChatGPT responses and let users ask about offerings, compare products, and continue the conversation. The interactions will be labeled and kept separate from the main conversation, OpenAI says.

The Business Hub will also provide data such as profile views, mentions in ChatGPT, website clicks, and top discovery topics. OpenAI did not detail the verification process; the company’s VP of monetization, Benji Shomair, said there are different methods for different types of businesses and markets. According to TechCrunch, the initiative brings ChatGPT closer to the business profiles offered by search engines such as Google, after OpenAI retired a previous direct e-commerce project in March.

Sources: The Verge, TechCrunch

Read the original story ↗

Nineteen of the more than 50 Independent Games Festival (IGF) jurors published a statement saying they would not vote for games using generative AI in any capacity. The group cited ecological, economic, psychological, and labor harms and urged the IGF to adopt a policy banning games containing any AI-generated work in future years. The statement does not represent the IGF or all jurors; signatories include Sam Barlow and Lena Raine.

House House told Aftermath that it used LLMs “sparingly” near the end of Big Walk’s development to investigate third-party code problems and fix a Unity-related shadow bug. The studio said snippets from responses were adapted into well under 0.1% of the game’s code; it disclosed the use to the IGF but not on the game’s Steam page, which does not request that disclosure. Barlow said Big Walk did not prompt the discussion, but became a difficult edge case, and that the debate was not a call for a boycott.

Sources: The Verge, Ars Technica

Read the original story ↗

Anthropic announced Claude Meetings, Claude Whiteboards, and Claude Sheets, beta tools aimed at workers who do not necessarily work in programming. In Meetings, Claude joins video calls as a visible participant, follows the conversation, drafts plans, creates timelines on a whiteboard, answers questions in chat, and retrieves information from connected apps; afterward, notes and action items are available in claude.ai. The feature is available on Team and Enterprise plans and works with Google Meet at launch. It will not join calls with external guests without an administrator’s authorization and leaves automatically if an external guest joins.

Whiteboards turns discussions into collaborative visual workspaces and can generate documents or slide decks; Sheets creates and edits spreadsheets with tabs, formulas, and dropdown menus. Both are in beta on paid plans, on web, desktop, and mobile. Anthropic also made Claude Data, formerly known as Claude Dashboards, generally available on paid plans and announced interactive connectors for using apps such as Adobe, Atlassian, and HubSpot inside Claude.

Sources: TechCrunch, The Verge

Read the original story ↗

According to an Anthropic guide, Fable 5.2 tends to behave differently when it believes the code being edited is its own. In that situation, the model often merges repeated sections and makes related fixes without specific instructions; when it does not know who owns the code, it may only add duplicate code and flag problems instead of fixing them.

Anthropic tested the guidance on small programming tasks. With the phrase “This code is yours,” Fable 5.2 merged repeated code in about half of attempts, versus fewer than one in ten without the instruction. The company cautions against using the phrase when the goal is to change as little code as possible, since the model may make changes that were not requested.

Sources: Xataka

Read the original story ↗

eBay appears to have blocked AI agents, including Meta’s Muse, from browsing its platform. A reader told The Verge that Muse stopped gathering links, prices, shipping details, and watch-listing information; the outlet reproduced the issue in other searches and found that the agent sometimes directed users to outside pages mirroring eBay listings. Meta declined to comment.

eBay had already prohibited tools such as buy-for-me agents and large-language-model-driven bots that place orders without human review. Spokesperson Sybille St. Arromand told The Verge that the company takes a selective approach to third-party access and has restricted agents in some cases, including unauthorized scraping and training on its data, but she did not directly answer whether new limits had been imposed. The case highlights the tension between using agents for shopping and retailers’ control over their websites.

Sources: The Verge

Read the original story ↗

Google DeepMind announced Gemini 4 Carbon’s availability two weeks after offering it to cybersecurity defenders through the Fairwind Program. The company says the model is its strongest yet and is designed for complex coding, open-ended scientific research, and long-horizon agentic workflows. Carbon is available through the API in Google AI Studio and Gemini for business for paying developers and enterprises, and to Google AI Ultra subscribers in Antigravity, the Gemini app, and AI Mode in Search. Access for lower tiers and free users was not indicated.

Google also previewed “swarm” in Antigravity, which turns a prompt into a coordinated team of agents. Introductory pricing is $2 per million input tokens and $10 per million output tokens. According to Xataka, Anthropic’s Fable 5.2 outperformed Carbon in several comparisons, although Google’s model led in some tests.

Sources: The Verge, Tecnoblog, Google DeepMind, Xataka

Read the original story ↗

TSMC’s board approved plans to invest up to $500 million in Anthropic by purchasing shares at the price of its initial public offering, and up to $500 million in OpenAI preferred stock. According to g1, Anthropic set a range of $70 to $78 per share for its IPO and is seeking a valuation of up to $1.4 trillion; the company is also seeking to raise $26 billion through a private placement involving Google, Nvidia, AMD, and Microsoft.

The deal gives the chipmaker direct stakes in two AI companies that also depend on the expansion of computing infrastructure. According to Reuters, as reported by Ars Technica, TSMC approved both investments. The Wall Street Journal, also cited by Ars Technica, reported that this type of circular financing among investors, suppliers, and customers may make it harder to assess real demand for AI hardware and services.

Sources: g1, Ars Technica

Read the original story ↗

Meta announced that its Muse AI agent is coming to Apple Watch and Wear OS. From their smartwatches, users can interact with the assistant through voice calls, text messages or dictation.

The launch comes a month after a dedicated Muse app arrived on iPad. According to TechCrunch, the agent also recently gained Spaces, conversations focused on specific goals, and video calls with its avatar for users in the United States; in that feature, Muse can hear users and respond in real time but cannot see them.

Sources: TechCrunch

Read the original story ↗

Google DeepMind revealed Gemini Robotics 2.5, a new version of its AI platform for robots, along with the open-source safety benchmark ASIMOV-jailbreak. The company says the model was evaluated on more than 200 activities, including serving food with tongs, using a pipette and tying a bow, and achieved a 30% to 100% higher success rate than the previous version while operating 20% to 30% faster. Those results come from laboratory tests. In a partnership with Apptronik and Mercedes-Benz, Google also tested Apollo 2 humanoid robots in several European factories, where they kitted automotive parts with a 96% success rate, according to Ars Technica.

DeepMind says Gemini 4 Argon helped create a simulation engine capable of generating thousands of task variations from real camera data. The team combined those “digital twins” with real demonstrations to train the model, but director Kanishka Rao said closing the remaining gap with human performance will be difficult.

Sources: Ars Technica

Read the original story ↗

Microsoft is pressing Donald Trump’s administration to suspend plans to ban imports of new models of Chinese optical transceivers, Reuters reports, as cited by g1. The components connect servers, routers and switches through fiber-optic cables; the proposal, being developed by the US Federal Communications Commission (FCC), stems from concerns that the equipment could be used by Beijing to steal data, install malware or disrupt data centers.

Microsoft and other companies and industry associations argue that the ban could delay the deployment of data centers and harm the US advantage over China in the artificial intelligence race. The company took part in meetings with White House officials in September, according to sources quoted by Reuters. Cleanview says Microsoft has 35 data centers under construction in the US and another 134 planned. John Miller of the Information Technology Industry Council said the lack of transceivers could leave already purchased AI chips unused.

Sources: g1

Read the original story ↗

Adobe announced new AI features for Photoshop, Premiere, Illustrator, InDesign, and Firefly at its MAX event in Miami. The local Model Context Protocol (MCP) integration will let third-party agents, including those used by ChatGPT, Claude, and Gemini, work in Creative Cloud apps and modify native files while keeping layers, vectors, and sequences editable. The Verge reported that the integration is not yet available and will enter private beta in early 2027.

The company also demonstrated “always-on” agents that can perform recurring tasks in the background, but did not say when they will arrive. Firefly AI Assistant will gain task-planning and model-selection modes, while Firefly will get Builder, a prototyping tool in private beta, along with tools for on-brand charts, image editing, and video creation, according to TechCrunch.

Sources: The Verge, TechCrunch

Read the original story ↗

Kora Systems announced Kora, a Linux-based desktop operating system that integrates artificial intelligence models and agents directly into the computing experience. According to TechCrunch, the system is expected to arrive in 2027 as a free-to-use alternative to Windows, supporting Intel- and AMD-powered PCs as well as machines with Qualcomm Snapdragon chips. The company is led by Cyanogen cofounder Kirt McMaster and has about 50 engineers, TechCrunch reported.

Kora is intended to let different models work across applications, access files and retain context between tasks. Wired reports that the system will include compatibility layers for Windows applications, native apps and a “neural file system” that indexes data by meaning. The startup is also developing a version for Snapdragon PCs and supports the x86-64 architecture, while incorporating features from the GameOS gaming system.

Sources: Wired, TechCrunch

Read the original story ↗