LectureRechercheRadarCadre d'investissement
Connexion / Inscription
ToutTechnologieProduitsInvestissementBourseAutres

Lecture

2026-10-29198 publications

Actualiser

Apple Adds 'Exclude Apps' Setting to iOS 27.2, Allowing Users to Block Siri AI Data Access

Sujet · Siri AIMatériel
2026-10-29 04:04 GMT+8

The Electronic Frontier Foundation (EFF) reported that Apple has introduced a new "Exclude Apps" setting in its latest iOS 27.2 update. This feature allows users to explicitly designate which applications' contents are inaccessible to Siri AI, addressing previous privacy concerns where the AI assistant might process data from end-to-end encrypted messaging apps.

Many device AI features rely on cloud processing, meaning that despite engineered privacy protections, content from secure messaging apps could leave a user's device without their knowledge. In its "Encrypt It Already" campaign earlier this year, EFF called for operating systems to provide straightforward controls allowing users to tell the system AI not to access specific apps, even if it meant losing some features.

According to EFF, controlling Siri AI's app access behavior was confusing in previous versions, often requiring users to toggle multiple options or disable Siri AI entirely and revert to Siri Classic. The new update simplifies this: users can navigate to Settings > Siri > Excluded Apps, tap "Add Apps," and select any apps they wish to shield.

The change does come with functional trade-offs. For instance, excluding the Messages app prevents Siri AI from reading or responding to messages. Similarly, excluding Signal means Siri cannot discern information within that app via on-screen awareness. EFF notes that these disabled features actually demonstrate the effectiveness of the new setting.

The article also compares Android's current state. While Google Gemini offers some controls, such as revoking access to connected apps, its interface is less intuitive than Apple's new settings and lacks per-app screen context controls. EFF suggests Google should further improve by offering stricter per-app toggles to reduce the risk of user error.

Sources :Electronic Frontier Foundation

Recherche

South Korean Telecom LG U+ Says On-Device AI Optimization Tech Accepted at NeurIPS, Claims 4.5x Speedup

Sujet · LG U+Prise rapide
2026-10-29 08:29 GMT+8

South Korean telecommunications operator LG U+ announced on October 29 that two of its research papers on artificial intelligence (AI) token optimization technology have been accepted for the "On-Device Intelligence" workshop at NeurIPS 2026. NeurIPS is one of the world's three top-tier AI academic conferences alongside ICML and ICLR, with this year's event scheduled for December in Sydney, Australia.

The research aims to address the constraints of limited memory and battery life in smartphones by improving the computational efficiency of tokens, the basic unit of data processed by AI, thereby reducing response latency and power consumption. LG U+ stated that it applied lightweighting and inference acceleration technologies to small language models (sLMs) used for call services. According to the company, processing speed increased by approximately 4.5 times, while battery consumption decreased by about 77% compared to existing methods.

Additionally, LG U+ developed technology for additional AI learning within the smartphone itself, reducing computation and memory usage to allow model training on the device without transmitting personal data to external servers. Verification on actual smartphones showed accuracy levels similar to server-trained models. Lee Sang-yeop, Chief Technology Officer (CTO) of LG U+, said the company is considering applying these findings to its AI services and plans to expand the optimization techniques validated on smartphones to server GPU environments.

Sources :zdnet.co.kr

Recherche

AI Engineer Podcast: Anthropic Scales Code Merges to 30,000 Daily via Claude Review

Sujet · AnthropicMatériel
2026-10-29 07:00 GMT+8

The AI Engineer podcast has published a new episode titled "How Anthropic Merges 30,000 Code Changes a Day," featuring Manish Kapur from Sonar and Addy Osmani from Anthropic. The discussion centers on the challenges posed by coding agents generating pull requests faster than traditional human review processes can handle.

According to the episode description, Anthropic scaled its internal software factory from about 2,500 to 30,000 merged changes per day. This increase was achieved by implementing Claude for first-pass reviews, supported by real verification evidence. The guests also discuss the concepts of "light" versus "dark" software factories and findings from Sonar's State of Code survey regarding where next efficiency gains might come from.

The episode covers technical controls such as permissions and sandboxes, and concludes with a lightning round on labeling agent-written PRs and determining when it is safe to merge. This resource offers a specific look at how a major AI organization is restructuring its engineering validation loops to accommodate high-throughput AI-generated code.

Sources :youtube.com

Recherche

South Korean AI Chip Startup FuriosaAI Completes 820 Billion KRW Series D Round at 3.8 Trillion KRW Valuation

Sujet · FuriosaAIMatériel
2026-10-29 08:21 GMT+8

South Korean artificial intelligence semiconductor startup FuriosaAI announced on October 29 that it has successfully completed a Series D funding round totaling 820 billion KRW. The final amount significantly exceeded the company's initial target range of 450 billion to 750 billion KRW, bringing its post-money valuation to approximately 3.8 trillion KRW.

The round was led by South Korea's government-backed National Growth Fund, which invested 370 billion KRW as the largest participant. Among overseas investors, Infinitum, which previously led major investments in Groq, made the largest contribution. Other global professional investment firms joining the shareholder base include QUDA Capital Partners, led by former Nvidia executive Charlie Fu, Cisco Investments, and Abu Dhabi-based Neoport Holdings.

Existing shareholders and domestic financial institutions also participated in follow-on investments. Naver continued its support through its Silicon Valley-based venture arm, while existing investors such as NextTrans, Korea Development Bank (KDB), DSC Investment, and Kakao Investment collectively added 170 billion KRW. Notably, the Export-Import Bank of Korea (KEXIM) selected FuriosaAI as the first venture company to receive direct investment since its establishment.

FuriosaAI plans to deploy the newly secured capital primarily to expand the global commercialization of its second-generation AI inference chip, 'RNGD' (Renegade), and to strengthen its software ecosystem. The RNGD chip has already been deployed in real-world enterprise and cloud environments, including LG AI Research, Samsung SDS, and Upstage's Daum search summarization service. Concurrently, the company is accelerating the development of its third-generation AI accelerator, designed for the hyperscale AI infrastructure market in collaboration with Broadcom.

Sources :zdnet.co.kr

Recherche

ServiceNow Expands Partnership with OpenAI to Integrate ChatGPT Plugin, Codex, and Joint Security Solutions into Enterprise Workflows

Matériel
Vérifié 2026-10-29 10:09 GMT+8

US enterprise software company ServiceNow (NYSE: NOW) announced on October 28, 2026, an expanded partnership with US artificial intelligence company OpenAI to help enterprises move seamlessly from AI reasoning to governed action. The expansion includes a new plugin allowing users to execute ServiceNow workflows directly within ChatGPT, plans to integrate OpenAI's Codex coding assistant into ServiceNow's AI Workflow Factory, and the development of joint cybersecurity solutions combining both companies' capabilities.

According to the official press release, the new ServiceNow plugin enables customers to complete governed and secure ServiceNow workflows through ChatGPT. Powered by ServiceNow Action Fabric, the integration combines ChatGPT's reasoning with ServiceNow's enterprise context, business rules, and permissions. IT teams can find similar incidents, assess change risks, and update records without leaving ChatGPT. ServiceNow also plans to preview richer interactive experiences, such as allowing on-call engineers to review alerts, coordinate responses, and update incident records within the chat interface.

In the development sector, the companies plan to integrate Codex with ServiceNow's AI Workflow Factory, helping developers build applications using ServiceNow's business context and platform capabilities. OpenAI models will also power Build Agent and Autonomous Engineer to support application modernization. On the security front, ServiceNow and OpenAI are collaborating on a joint solution that combines OpenAI's cybersecurity capabilities with ServiceNow's asset, identity, and enterprise context to help teams identify, triage, validate, and mitigate attack paths.

Internal operations are also shifting: OpenAI plans to use ServiceNow across its own operations, while ServiceNow intends to make ChatGPT Work and Codex available to all employees and use OpenAI capabilities in its internal security work. Regarding availability, the ServiceNow MCP plugin is live in the OpenAI marketplace today, with enhanced capabilities expected by end-of-2026. Voice features in the Super Intelligence Control Tower are slated for Q1 2027, while OpenAI-powered security capabilities are expected to enter early access in January 2027, with general availability planned for Q2 2027.

Sources :OpenAI

Recherche

Samsung Electronics Announces 84.4 Trillion KRW Capex Plan for 2026, Up 67.8% Year-on-Year

Sujet · 三星电子2026资本支出Matériel
2026-10-29 09:27 GMT+8

South Korea's Samsung Electronics announced on October 29, 2026, that it will spend 84.4 trillion KRW (approximately $59 billion) on facility investments this year, marking a 67.8% increase from last year's total of 50.3 trillion KRW. This significant expansion is primarily driven by artificial intelligence demand, aiming to strengthen its position in the global semiconductor supply chain.

In terms of departmental allocation, the Device Solutions (DS) division, which handles semiconductors, will receive 76.8 trillion KRW, while the Display division will get 3.9 trillion KRW. Samsung stated that the memory business will focus on investing in new fabs and infrastructure to address robust AI demand and long-term market changes. The foundry business will center its investments around the new factory in Taylor, Texas, to rapidly respond to surging customer demand for AI and High-Performance Computing (HPC).

The substantial rise in capital expenditure reflects continued heavy investment by global tech giants in AI computing infrastructure. As one of the world's largest memory manufacturers, Samsung is expanding capacity to consolidate its competitive advantage in High-Bandwidth Memory (HBM) and advanced logic chips. Additionally, Samsung disclosed R&D expenses of 14.4 trillion KRW for the third quarter, indicating sustained investment in future technological competitiveness.

Sources :미래를 보는 창 - 전자신문

Recherche

US DOJ Investigates Nvidia's $12.9 Billion Acquisition of Hugging Face

Matériel
2026-10-29 04:13 GMT+8

According to CNBC, the U.S. Department of Justice (DOJ) is investigating chipmaker Nvidia and has requested information on how its $12.9 billion acquisition of Hugging Face, an AI model hosting and testing platform, could impact competitors. Reuters first reported the inquiry on Wednesday, which CNBC confirmed with a person familiar with the matter.

Hugging Face runs a widely used open-source platform for hosting and testing AI models, while also selling infrastructure and collaboration tools to the tech industry. When Nvidia announced the acquisition in September, CEO Jensen Huang stated the deal would help "expand access to AI for developers and institutions worldwide." At that time, Nvidia said more than 18 million people and over 200,000 companies used Hugging Face to share at least 3 million models and 500,000 datasets.

This acquisition marks Nvidia's second-largest purchase, following its $20 billion acquisition of Groq assets late last year, which also attracted DOJ scrutiny. The closest comparable to Hugging Face's business was GitHub, acquired by Microsoft in 2018 for $7.5 billion.

Nvidia declined to comment. The investigation is currently in the preliminary information-gathering stage, with no specific legal charges or next procedural steps disclosed yet.

Sources :CNBC

Recherche

Zillow Agrees to $212.5 Million Settlement in Investor Class Action Over Failed Home-Flipping Venture

Sujet · ZillowMatériel
2026-10-29 08:03 GMT+8

Online real estate marketplace Zillow announced on October 28, 2026, that it has agreed to pay $212.5 million in cash to settle a class action lawsuit brought by investors. The suit alleged that company leaders misled shareholders about the risks associated with Zillow Offers, the firm's now-defunct home-flipping venture.

The settlement, known as the Jaeger case after lead plaintiff Jeremy Jaeger, was filed in the U.S. District Court for the Western District of Washington and disclosed in an 8-K form to the U.S. Securities and Exchange Commission. The plaintiffs accused Zillow of inflating its stock price by failing to disclose the unpredictability of its home price forecasts, which contributed to the collapse of the iBuyer program.

This resolution comes shortly after the U.S. Supreme Court declined to hear Zillow's petition to review the lower court's class certification ruling. Steve Berman, managing partner of Hagens Berman, the court-appointed class counsel, described the outcome as likely the largest single securities settlement in the history of the Western District of Washington. A spokesperson for Zillow stated the company is pleased to reach a resolution.

Under the terms, Zillow will deposit the funds into a settlement account within 20 business days of preliminary court approval. The class includes investors who purchased Zillow's Class A or Class C stock between August 5, 2021, and November 2, 2021. A period between preliminary and final approval hearings will allow class members to opt out or object.

Sources :RealEstateNews.com

Recherche

Alphabet Reports Rare Top-and-Bottom Line Miss as AI Capex Drives Free Cash Flow Negative

Sujet · AlphabetMatériel
2026-10-29 08:48 GMT+8

U.S. tech giant Alphabet (Google's parent company) released its financial results for the third quarter of fiscal year 2026, reporting that both revenue and adjusted earnings per share fell short of Wall Street expectations. This marks the company's first simultaneous miss on both lines in nearly four years. Shares of Alphabet dropped more than 9% in after-hours trading following the announcement.

Specifically, Alphabet's Q3 revenue rose 25% year-over-year to $127.58 billion, but this still trailed the analyst consensus of $128.22 billion. Adjusted EPS came in at $2.67, significantly below the expected $3.02. More concerning for investors was the company's free cash flow, which recorded a negative $9.07 billion, far below the projected positive $2.7 billion.

Although Chief Financial Officer Anat Ashkenazi reiterated the full-year capital expenditure guidance of $195 billion to $205 billion, with projections exceeding $300 billion for fiscal 2027, this quarter's performance indicates that the massive investment in artificial intelligence infrastructure is no longer just impacting cash flow but is directly compressing operating profits. Quarterly capex totaled $51.36 billion, slightly below the expected $55.09 billion, yet net income declined from $34.98 billion a year ago to $26.17 billion.

Notably, Alphabet's cloud business maintained strong momentum. Google Cloud revenue doubled to $30.7 billion, surpassing the analyst target of $28.7 billion. CEO Sundar Pichai stated that the Gemini app reached over one billion users globally for the first time, becoming the fastest-growing product in the company's history. However, as Investing.com analyst Thomas Monteiro pointed out, Alphabet currently earns more from selling AI compute to rivals than it does from using it itself, a structural tension that will continue to pressure margins as depreciation rises and the cost of capital is repriced upward.

Sources :SiliconANGLE

Recherche

Japanese Motor Giant Nidec Delays Resolution of Audit Disclaimer as PwC Demands Investigation of 132 Executives

Sujet · 尼得克审计争议Matériel
2026-10-29 17:00 GMT+8

Japanese motor manufacturer Nidec announced on October 28, 2026, that it could not resolve the audit opinion disclaimer for its fiscal year ending March 2026 by the planned deadline. The company had aimed to obtain an audit opinion and submit corrected securities reports by this date, but its accounting auditor, PwC Japan, requested new additional investigations.

According to EE Times Japan, PwC detected matters requiring further confirmation during email screening procedures, leading to an expansion of the investigation scope. Specifically, regarding "Management Representation Letters" submitted by heads of various business bases to auditors, PwC requested additional investigation of 252 signature candidates. After consultation, 120 individuals were assessed based on existing findings, while the remaining 132 require verification through email confirmations, interviews, and external expert evaluations.

Nidec plans to complete the additional work for these 132 individuals within November 2026, but the overall audit completion timeline remains undetermined. This delay means Nidec will remain designated as a security under special caution by the Tokyo Stock Exchange. The company has received audit disclaimers for two consecutive fiscal years (ending March 2025 and March 2026), primarily due to incomplete identification and personnel correction measures for employees involved in improper accounting practices, as well as insufficient audit evidence for accounting treatments such as asset impairment losses.

Sources :EE Times Japan

Recherche

Semiconductor Equipment Company KLA Reports Record Quarterly Revenue, Forecasts Over 70% Growth in Advanced Packaging Equipment Sales

Sujet · KLA业绩Matériel
2026-10-29 08:04 GMT+8

US-based semiconductor inspection and metrology company KLA (NASDAQ: KLAC) reported its first-quarter fiscal 2026 results ending in September on October 28, 2026. Revenue reached a record $4.14 billion, up 29% year-over-year and 13% sequentially. The operating margin rose to a company-record 46.1%, and earnings per share exceeded the midpoint of previous guidance.

CEO Rick Wallace attributed the growth to strong demand for process-control equipment tied to artificial-intelligence infrastructure investments, particularly in leading-edge logic, memory, and advanced packaging. KLA expects revenue from advanced-packaging process-control systems to reach approximately $1.1 billion in calendar 2026, representing more than 70% year-over-year growth. This rate is nearly twice the expected growth of the broader advanced-packaging market.

Looking ahead, KLA projected December-quarter revenue of about $4.7 billion. The company also raised its estimate for the calendar 2026 wafer-equipment market (WFE) to the mid-$150 billion range and anticipates the market will exceed $200 billion in 2027 as AI-driven expansion continues.

Despite absolute growth, CFO Bren Higgins noted that China's share of total revenue is expected to decline to the low-20% range in 2027, as spending outside China grows more rapidly. Currently, China accounts for roughly 27% to 29% of the expected fourth-quarter revenue.

Sources :MarketBeat

Recherche

Sonar Engineer Taylor Luttrell-Williams Demos Hunter Agent: How AI Identifies Logic Bugs Missed by Traditional SAST

Matériel
2026-10-29 03:00 GMT+8

Traditional pattern-based Static Application Security Testing (SAST) excels at catching bad code but often fails to identify valid code with flawed intent, such as broken access controls, business logic gaps, and authentication holes. Addressing this limitation, Taylor Luttrell-Williams of Sonar introduced the SonarQube Hunter Agent during a presentation at the AIE CODE summit.

According to the episode description, the Hunter Agent operates not through simple rule matching, but by mapping the entire repository, running three specialized AI agents for deep analysis, and adversarially validating every finding. This design aims to simulate attacker thought processes to uncover logical flaws that human developers or traditional tools might overlook.

During the demonstration, Luttrell-Williams showcased the tool's performance on a Python application. The Hunter Agent successfully identified 46 security issues, including a vulnerability allowing regular users to escalate privileges to admin status and an MFA bypass. These cases concretely illustrate the potential advantages of AI agents in handling complex logic vulnerabilities.

The talk also noted that Sonar has joined OpenAI's Daybreak cyber partner program, marking a significant step in leveraging advanced AI models to enhance cybersecurity defenses. For professional readers interested in practical applications of AI in software engineering and security, this video offers a direct reference on how agentic workflows are changing code auditing.

Sources :youtube.com

Recherche
Suivant →
Connexion / Inscription
LectureRechercheRadarCadre d'investissement