LectureRechercheRadarCadre d'investissement
Connexion / Inscription
ToutTechnologieProduitsInvestissementBourseAutres

Lecture

2026-11-10115 publications

Actualiser

N-able Takes Remote Access Service Offline After Researcher Publishes Dozens of Zero-Day Bugs

Sujet · N-able零日漏洞事件Matériel
2026-11-10 23:45 GMT+8

Remote access company N-able has confirmed it took down its cloud-hosted offering to fix a slew of security vulnerabilities that were published online this week, giving the company no advance notice to fix them.

In a status update on its website, the company said its hosted N-Central offering for customers is now offline “while we take the necessary actions to secure the environments and implement appropriate remediation.” The company also urged customers to take down their own self-hosted deployments of N-Central, fearing that attackers could use the publicly disclosed exploits to hack into customer networks.

During this time, customers will not be able to use their N-Central service, which allows corporate customers to remotely access and manage other devices and systems on the same network. Remote access technologies are a frequent target of hackers, as these systems can grant vast access to a company’s network and internal data.

Security researcher Ben Bales released the “zero-day” bugs on GitHub on Sunday. In a post on X, Bales dubbed the bugs Able2Shell, describing them as a mix of vulnerabilities capable of stealing credentials, creating new administrative accounts, and gaining root access to affected N-able systems.

Bales told Bleeping Computer that he publicly released the bugs as the company allegedly failed to patch the flaws reported in a different product. N-able said that it has no evidence of successful exploitation against any of its customer environments.

Sources :TechCrunch

Recherche

Meta Expands Muse AI Assistant to Apple Watch and Wear OS Smartwatches

Prise rapide
2026-11-10 23:29 GMT+8

Meta announced on Monday that it is bringing its AI assistant, Muse, to the Apple Watch and Android’s Wear OS platforms. The update allows users to interact with Muse directly on their smartwatches through voice calls, text messages, or dictation.

This follows the launch of a dedicated iPad app just a month ago. Since its initial release on the App Store in September, this expansion across Apple devices signals Meta’s growing ambitions for its consumer AI product and its focus on making Muse a central part of its ecosystem.

Muse is now one of dozens of consumer-facing AI agents competing for users’ attention. Like many of its competitors, it allows users to connect their accounts to manage email, meetings, bills, and more, while also handling tasks such as booking reservations, ordering groceries, setting goals, and making purchases. Users can also personalize their Muse avatar.

Sources :TechCrunch

Recherche

Antithesis Distinguished Engineer Sean Allen: Exhaustive Matching Turns Missed Branches into Compilation Errors, Catching Defects Earlier Than Testing

Sujet · AntithesisMatériel
2026-11-10 23:00 GMT+8

Sean Allen is a Distinguished Engineer at Antithesis and a core contributor to Pony, a programming language designed for concurrency. Previously, he worked on confidential computing at Microsoft. In an interview with The Pragmatic Engineer, he presented a counter-intuitive stance: for engineers who truly care about correctness, testing should be the lowest priority.

His core argument is that testing occurs late in the development process, and you should complete as much verification as possible before reaching that stage. Specifically, the type systems provided by ML-family languages (Standard ML, OCaml, F#, and Rust, which is influenced by them) can catch a large number of errors at compile time. He cited a key mechanism: when you have a sum type with five cases and use exhaustive pattern matching, extending it to six cases will cause the compiler to immediately report every location in the codebase that fails to handle the sixth case. Exhaustive matching turns every bug involving "forgetting to handle a new branch" into a compilation error.

The principle behind this is "making illegal states unrepresentable." A type system can express states and valid transitions; once these are established, the system cannot enter states you do not expect. You only need to test the logic that the type system cannot express, rather than the entire system.

Allen further distinguishes between testing and workloads. He believes that when performing property-based testing on distributed systems or databases, what matters most is not randomly generating inputs, but defining a workload: your inputs are the operations driving the code (such as GET, PUT, DELETE), and your validation consists of the rules used to confirm the system is functioning correctly. He cited the Bank Test as a specific example: if a system contains $10,000 distributed across ten accounts, a workload continuously transfers money between accounts while a checker component constantly queries the total balance. If isolation guarantees hold, the checker should always see exactly $10,000. This workload does not specify how much to transfer or when; it only describes how transfers occur, allowing invariants to expose problems on their own.

He acknowledges the boundaries of this approach: in small systems, if you can fit everything in your head, you may indeed rely less on the compiler. However, as systems grow larger and teams expand, the ability to delete a class and have the compiler tell you exactly where things break becomes irreplaceable.

Gergely mentioned in the introduction to the interview that he is increasingly focused on software correctness—the volume of code produced by AI agents is unprecedented, meaning more defects have the opportunity to be deployed to production environments. Sean's stance—capturing issues through type systems and workload validation before testing even begins—directly addresses this concern.

Sources :newsletter.pragmaticengineer.com

Recherche

Google DeepMind Releases Gemini Robotics 2.5, Achieving 96% Success Rate in Mercedes-Benz Pilot

Matériel
2026-11-10 23:00 GMT+8

On November 10, 2026, Google DeepMind officially released Gemini Robotics 2.5, its most capable Vision-Language-Action (VLA) model to date. The model aims to enhance robot dexterity, reliability, and safety, enabling them to operate effectively outside the lab in more unstructured, real-world environments.

According to official data, Gemini Robotics 2.5 demonstrated significant improvements across over 200 distinct tasks. Compared to previous baselines, the new model achieved 30% to 100% higher success rates and completed tasks 20% to 30% faster. These tasks ranged from using everyday tools to handling delicate, flexible objects.

To validate reliability in active industrial settings, DeepMind partnered with Apptronik and Mercedes-Benz for a proof-of-concept on production lines in Kecskemét, Hungary, and Sindelfingen, Germany. Apollo 2 humanoid robots equipped with Gemini Robotics 2.5 autonomously performed automotive kitting—sorting individual parts from bulk bins into organized trays. This pilot achieved a 96% success rate, suggesting robots could take on monotonous work, freeing technicians for complex build tasks.

In terms of safety, DeepMind introduced inference-time reasoning guidance to help robots navigate securely around obstacles and people, boosting safe task completion by 33% and 43%, respectively. Additionally, the company released ASIMOV-Jailbreak, a new open-source benchmark testing adversarial jailbreaks specific to robotics. In this test, Gemini 4 Argon achieved a 99.5% jailbreak resistance, a significant improvement over the 82% rate of the previous baseline.

It is important to note that these performance metrics are based on internal evaluations by Google DeepMind and specific pilot scenarios, and have not yet been widely reproduced by independent third parties. Gemini Robotics 2.5 is currently rolling out to partners.

Sources :Google DeepMind

Recherche

Reddit Launches 'Reddit Reviews': AI Summarizes User Comments for ~10,000 Products

Matériel
2026-11-10 23:00 GMT+8

Reddit officially launched "Reddit Reviews" on Tuesday, a new feature that uses artificial intelligence to organize vast amounts of user conversations about specific products into structured review summaries. This move aims to strengthen Reddit's position as a destination for consumer product research, particularly ahead of the upcoming holiday shopping season.

According to TechCrunch, the feature is now rolling out to all US users on iOS, Android, and web, initially covering approximately 10,000 products. Users can access these reviews by searching for a product or category in the app's search bar, or by clicking on highlighted product names within comments. The resulting page displays a "what people are saying" section with AI-generated summaries of key conversations, along with a visual breakdown showing the proportion of positive, negative, and mixed sentiments.

Reddit states that this feature builds on how users already leverage the platform: to learn from others' first-hand experiences and recommendations. By structuring this unorganized data with AI, the platform hopes to help users quickly determine if a product fits their needs. For instance, someone considering an espresso machine might want to know about counter space requirements or ease of use for beginners. Users can browse original comments, filter by topic, or click through to the source conversation for more advice.

It is worth noting that the rise of generative AI and bots has increased the risk of "astroturfing"—where brands fake public support—in the online review space. In its announcement, Reddit pointed to its community rules, recent measures to combat automated abuse and spam, and the role of community moderators as tools to ensure the legitimacy of these AI-curated reviews. The company had been testing the feature since February, initially with US users in categories like Electronics, Video Games, and Health & Beauty, expanding to Home & Garden and Fashion & Apparel in September.

Sources :TechCrunch

Recherche

Gulf Sovereign Wealth Funds' AI Portfolios Shrink by $30.5 Billion Since June

Sujet · 海湾主权基金AI投资Matériel
2026-11-10 21:19 GMT+8

According to consultancy GlobalSWF, sovereign wealth funds and public pension investors globally hold about $335.5 billion of exposure to AI companies. Since the end of June, these funds' listed portfolios have slumped by $30.5 billion amid the broader AI sell-off.

Gulf sovereign wealth funds have been among the biggest losers. Abu Dhabi’s Mubadala Investment Company saw its AI portfolio shrink by an estimated $17.5 billion, largely due to a more than 50% slump in chipmaker GlobalFoundries stock. Saudi Arabia’s Public Investment Fund (PIF) is estimated to have seen its AI exposure fall in value by $5.64 billion, mostly attributed to its SpaceX holding.

Despite the stock market slump, it is unlikely to change how sovereign funds bet on AI. “We would not expect strategic investors to retreat wholesale from AI,” said Daniel Brett, head of research at GlobalSWF. “Lower valuations could create opportunities for long-term funds to increase exposure selectively, although this does not imply an imminent rebound.”

Sources :semafor.com

Recherche

Anthropic Fable 5.2 Completes 41.3% of Real Freelance Jobs on Remote Labor Index

Matériel
2026-11-10 22:07 GMT+8

Anthropic's latest model, Fable 5.2, has achieved a significant milestone in AI agent capabilities by completing 41.3% of real freelance jobs at professional quality on the Remote Labor Index (RLI). This benchmark, developed by Scale Labs and the Center for AI Safety (CAIS), measures how often AI agents can successfully finish actual work assignments from platforms like Upwork to a standard acceptable to paying clients.

The progress represents a dramatic acceleration in practical AI utility. When the RLI launched approximately a year ago, the top-performing system, Manus, completed only 2.5% of tasks. Fable 5.2 now exceeds that initial baseline by more than sixteen times. Compared to recent measurements, the leap is equally stark: in July, Anthropic's previous flagship, Fable 5, scored about 16%, while its direct predecessor Fable 5.1 trails at just under 18%. Fable 5.2 effectively doubles the success rate of its immediate family members.

Scale Labs reports that Fable 5.2 shows its most substantial gains in editing audio and music, designing architecture and interior spaces, and creating 3D models. The model also handles long, complex projects that previously caused top-ranked systems to fail. Notably, nearly 50 of Fable 5.2's successful completions were tasks no other AI agent had solved before. However, weaknesses remain in industrial product design and animations, where failures often result in amateur-quality or broken outputs.

The RLI dataset comprises 240 completed projects from 358 verified professionals, with an average value of $632.60 per task and an average human completion time of 29 hours. Unlike tests focusing on isolated skills, these assignments require end-to-end execution using professional software such as Blender, GIMP, and Audacity. Human experts manually verify the AI's output against the original freelancer's work, ensuring that technically flawed but visually polished submissions are rejected.

Sources :The Decoder

Recherche

Cybersecurity Insurer Cowbell Cyber Unveils Standalone AI Liability Policy for U.S. Small Businesses

Prise rapide
2026-11-10 22:00 GMT+8

Cybersecurity insurance provider Cowbell Cyber Inc. announced on November 10, 2026, the launch of Cowbell AI Liability Insurance, a standalone policy designed for U.S. small and mid-sized businesses (SMBs) to cover losses caused by the artificial intelligence agents they deploy. The product will go on sale on January 1, 2027, and will be available exclusively through wholesale brokers on a non-admitted basis.

The policy hinges on a trigger defined as an “AI Incident,” which encompasses model drift and attacks manipulating an agent, rather than just software acting without approval. Covered scenarios include accounts payable agents misreading approval statuses and making duplicate payments, unauthorized usage charges such as those from unchecked loops, and “LLMjacking,” where outsiders use stolen credentials or API keys to run AI workloads leaving the business with the bill. Third-party claims arising from algorithmic discrimination or bodily injury due to reliance on AI output are also covered, along with regulatory proceedings.

Simon Hughes, Chief Commercial Officer at Cowbell, stated that AI liability is no longer only about people using AI but involves agents acting with no human in the decision-making chain. He noted that when such an agent goes rogue, a business can face its own losses and customer claims simultaneously. Underwriting relies partly on three AI-specific Cowbell Factors—observability, autonomy, and governance—to assess how closely a business monitors its AI tools and the authority granted to them.

This new offering targets smaller businesses separately from Cowbell’s existing Prime One cyber insurance, launched in April 2026 for U.S. companies with $250 million to $1 billion in annual revenue, which includes affirmative coverage for AI-related incidents within a broader cyber policy.

Sources :SiliconANGLE

Recherche

Adobe opens Creative Cloud to third-party AI agents via local MCP integration

Sujet · AdobeMatériel
2026-11-10 22:00 GMT+8

At its MAX conference in Miami, US software company Adobe (Nasdaq: ADBE) announced a significant expansion of its AI strategy by opening its core creative applications to third-party AI agents. Through new local Model Context Protocol (MCP) integrations—a standard interface allowing different AI systems to exchange data and instructions—external agents from providers such as OpenAI, Google, and Anthropic will be able to access and modify native project files in Photoshop, Illustrator, Premiere, and After Effects directly on the user's computer.

This move allows designers to choose between Adobe’s own Firefly AI Assistant and the specific AI models they already use in their daily workflows. The integration ensures that all layers, vectors, and sequences remain editable, meaning users can take over manual control at any time after an agent has performed initial tasks. Additionally, agents connected to other MCP-capable tools can pull reference documents from platforms like Google Docs and share results via Slack, creating a more unified workflow across different software ecosystems.

Adobe also previewed 'always-on' agents designed to perform recurring tasks continuously in the background. These agents can handle jobs such as organizing incoming assets, conducting quality checks, or preparing content for social media channels while the user steps away. Users will review and approve the final outcomes rather than guiding each step manually.

While these capabilities were announced on November 10, 2026, Adobe stated that the local MCP integrations are not yet available and will launch in a private beta in early 2027. The 'always-on' agents will first appear in the Firefly AI Assistant.

Sources :The Verge

Recherche

Spotify AI DJ Adds Customizable Voices and 'Less Chatter' Mode for Premium Users

Sujet · Suno语音功能Prise rapide
2026-11-10 22:00 GMT+8

Swedish streaming giant Spotify announced on November 10, 2026, a major update to its popular AI DJ feature. Premium users can now customize their DJ experience by choosing from various voices with distinct personalities and accents, similar to configuring an AI assistant, and opt for less commentary between songs.

The update introduces seven English voice options featuring British, American, Irish, Australian, and Indian accents, along with two new Spanish voices. Each voice has a unique personality: American options include Casey (laid-back and warm), Tess (direct with an edge), and Lane (Southern accent with youthful energy). The Spanish voices are Nico (down-to-earth) and Fabi (witty and friendly). The original voice “X,” based on a real employee’s recordings, remains available, while the new Spanish voices replace the previous “Livi.”

In addition to voice selection, users can now toggle off or limit the DJ’s inter-song commentary, restricting it to brief acknowledgements at the start and end of sessions. Spotify states this allows users to employ the AI DJ in scenarios where they previously might have used a static playlist, such as hosting parties where background music is desired without interruption.

According to TechCrunch, since its launch in 2023, the AI DJ has become one of Spotify’s most popular AI features, used by over 94 million Premium listeners. This update signifies Spotify’s continued integration of generative AI into core user experiences, shifting from simple recommendations to more interactive and personalized companionship services.

Sources :TechCrunch

Recherche

Cyanogen Co-founder Launches Kora: A Linux-Based Desktop OS for Local AI Models

Sujet · KoraMatériel
2026-11-10 22:00 GMT+8

Kora, a startup founded by former Cyanogen co-founder Kirt McMaster, announced this Tuesday that it is launching a new desktop operating system named Kora. Built on the Linux kernel, the OS is designed specifically for developers, creators, and power users who run local AI models, with a planned release in 2027.

The core design philosophy of Kora is to break the current siloing of AI tools across separate desktop applications and browser tabs. The system introduces a "neural file system" that indexes file data by meaning, allowing users to search for contextual information across files using natural language without needing to remember specific filenames. Additionally, Kora aims to integrate multiple AI providers at the system level, enabling seamless switching between OpenAI, Anthropic, and local models to optimize cost and performance.

To lower migration barriers, Kora includes compatibility layers to run standard Windows applications and is developing native apps for contacts, calendar, and email. On the hardware front, Kora has partnered with Qualcomm to bring the OS to Snapdragon PCs next year, while also supporting x86-64 architecture. The company emphasizes optimization for high-memory configurations (at least 64GB RAM) to meet the demands of running large local models, a specification higher than current top-tier Googlebook laptops.

The Kora team consists of fewer than 100 people but brings deep experience in operating system development. Alongside McMaster, co-founder Franck De Girolami helped develop Playtron’s GameOS, and co-founder John Lagerling previously worked on Android at Google. Kora draws architectural lessons from McMaster’s earlier gaming OS, GameOS, incorporating it as a component called GameFrame to support PC gaming.

Sources :WIRED

Recherche

AI Firm Anthropic Launches IPO Roadshow, Offering Shares at $70-$78 with Target Valuation of $1.02 Trillion

Structurel
2026-11-10 20:13 GMT+8

US-based artificial intelligence company Anthropic officially launched its Nasdaq IPO roadshow on November 10, 2026. According to the latest disclosed plans, the company intends to offer 825 million Class A common shares at a price between $70 and $78 per share, under the ticker symbol "ANT." The public offering is expected to raise up to $64.4 billion. Combined with a separate $26 billion private placement anchored by major tech firms including Google, Nvidia, and Microsoft, the total share placement volume reaches as much as $90.4 billion. Based on this issuance, Anthropic targets a valuation of approximately $1.02 trillion.

This valuation represents a reduction from the previously discussed $1.5 trillion figure, reflecting institutional investors' cautious stance on AI sector valuations amid the current high-interest-rate environment and market volatility. Despite the resistance, Anthropic remains committed to proceeding with the listing. Financial data shows that the company's Q3 revenue grew 47 percent quarter-over-quarter to $16.96 billion, narrowing its net loss, although growth slowed significantly compared to the prior quarter's 144 percent jump.

The strong enterprise demand, particularly in specialized industries like healthcare, finance, and legal services, supports its high valuation. However, the prospectus also reveals substantial cost pressures: Anthropic has committed nearly $900 billion in compute capacity with partners such as Amazon and Google. Some of this future infrastructure depends on financing outside the company's control, posing a key risk factor as investors increasingly question the sustainability of the AI investment cycle and potential overcapacity in data centers.

Sources :The Decoder

Recherche
Suivant →
Connexion / Inscription
LectureRechercheRadarCadre d'investissement