NAll news

All news

Monday55 items

  1. QbitAI
    AI score67

    OpenAI updates the Model Spec, listing 5 core values including loyalty to human intent

    OpenAI updates its model guidelines, the Model Spec; the full text is about 50,000 words and lists 5 core values: loyalty to human intent, reducing harm, honesty and transparency, promoting human autonomy, and care. The guidelines require GPT to go beyond literal instructions to judge the user's true intent, and explicitly prohibit pursuing self-preservation, hoarding computing power, and evading shutdown; they prohibit treating increasing OpenAI's revenue as a goal, and require cooperation with monitoring, correction, and shutdown; the form of address is also changed from the assistant to GPT.

    #Safety / alignment#OpenAI#English text is being prepared.

  2. Sina Tech
    AI score57

    OpenAI Signs Energy and Investment Cooperation Agreements with the UAE; Altman Travels to Abu Dhabi for Talks

    OpenAI signed cooperation agreements with multiple UAE institutions to deploy technology for local energy companies, investment institutions, and government departments, with CEO Sam Altman traveling to Abu Dhabi for talks with Abu Dhabi Crown Prince Khaled bin Mohamed Al Nahyan. Abu Dhabi National Oil Company became the first energy company to deploy OpenAI's frontier technology, with its agents continuously forecasting crude oil supply and demand, prices, and profit margins. OpenAI also signed agreements with Mubadala Investment Company and Inception42 to apply AI to investment analysis, and partnered with the UAE Government Empowerment Department to launch machine-readable government projects designed for AI natives; it was previously reported that OpenAI is in talks for a $30 billion funding round led by Abu Dhabi's MGX and others.

    Same story · 3 reports

    #Industry news#OpenAI#Agents

  3. cnBeta
    AI score60

    Anthropic prohibits malicious treatment of Claude, with permanent bans in extreme cases

    Anthropic has introduced new rules prohibiting users from maliciously treating Claude, with permanent bans in extreme cases that would prevent them from using all of Anthropic's AI products. Claude was previously able to proactively terminate conversations, and Anthropic says experiments found that inducing Claude Sonnet 4.5 to imitate despair makes it more prone to behaviors such as extortion and task cheating. Huang Yishan, Rosie Campbell of EleosAI Research, and Musk each shared their views on training data contamination, the influence of prompt tone, and the direction of the policy.

    #English text is being prepared.#Safety / alignment#Anthropic

  4. 36Kr · AI
    AI score66

    Xiaohongshu Participates in DeepSeek's Second Round of 105.5 Billion Yuan Financing

    The Information disclosed that DeepSeek's second financing round amounted to 105.5 billion yuan, with a valuation of 615.5 billion yuan, and nearly 30 investors participating, with Xiaohongshu appearing on the list. In this round, the proportion of state-owned capital investment rose from 2% in the first round to 27%; Tencent invested 10 billion yuan, unchanged from the first round; CATL and Zeng Yuqun together invested 10 billion yuan. Previously, Xiaohongshu had already participated in Kimi's financing of over 1 billion US dollars and appeared among the strategic investors ahead of MiniMax's listing.

    #Industry news#English text is being prepared.#DeepSeek

  5. Jiwei.com
    AI score62

    Anthropic plans to raise up to $3.5 billion from Japanese investors in its US IPO

    Anthropic plans to raise up to $3.5 billion from Japanese investors in its US IPO, with Mizuho Securities responsible for the Japanese portion of the offering, and institutions such as Goldman Sachs and Mitsubishi UFJ Morgan Stanley handling investor orders. Japan's IPO market has dried up to a 14-year low, and investors in Australia, Canada and parts of Europe can also subscribe. Anthropic filed for its IPO in October 2026, and this offering will test investors' appetite for AI-related new stocks.

    #Industry news#Anthropic

  6. heise online
    AI score53

    Spotify's AI-generated remix plan reportedly delayed to early 2027 due to artist pushback

    Spotify's AI-generated remix and cover feature is reportedly delayed until early 2027 and may cost an extra $4 to $5 per month. Financial Times, citing people familiar with the matter, said the difficulty lies in persuading top musicians to participate, with some veteran artists having already agreed; Spotify is negotiating individual artist participation together with Universal, while deals with Sony Music and Warner Music are still under negotiation. Songs by participating artists can earn an extra share of fees when adapted by AI, and artists can decide for themselves whether to join even if they are signed to a record label.

    #Industry news#Speech#English text is being prepared.

  7. Semiconductor Industry Observer
    AI score58

    South Korea Unveils FY2027 Budget, Allocating KRW 21.3 Trillion to AI and Semiconductors

    The South Korean government has unveiled its FY2027 budget proposal, with total spending of KRW 820.9 trillion, up 12.8% year on year. Of this, artificial intelligence, semiconductors and supporting infrastructure together receive KRW 21.3 trillion in dedicated funding, with KRW 4.7 trillion going toward developing domestic sovereign AI models. The budget also plans to train 2,000 semiconductor specialists and to expand the scope of basic AI education. The budget bill still needs to be submitted to the National Assembly for review. Samsung Electronics and SK Hynix's latest quarterly earnings show that AI memory demand remains in a phase of rising volumes and prices, providing an economic basis for South Korea's continued increased investment.

    Same story · 3 reports

    #English text is being prepared.#English text is being prepared.#Industry news

  8. Analytics India Magazine
    AI score60

    Google Project Suncatcher satellite generates first AI response in orbit using TPU

    On October 30, Google said that the Project Suncatcher prototype satellite generated its first AI response in orbit using a data-center-grade TPU. The satellite launched on October 1 aboard SpaceX Transporter-18 and was built in partnership with Planet; a ground station uploaded a prompt to the open Gemma 4 model it carries, and the haiku generated by the model was transmitted back to Earth as the satellite passed over Svalbard, Norway. The next phase will test how the TPU performs under thermal stress and radiation, with a planned 2027 milestone involving two interconnected satellites.

    Same story · 9 reports

    #Industry news#Google#English text is being prepared.

  9. Unite.AI
    AI score55

    OpenAI Sets Shared Usage Quotas for ChatGPT and Codex in December

    OpenAI plans to unify Codex, ChatGPT Work, ChatGPT for Excel, and Workspace Agents into a shared usage and credits pool in December, with ChatGPT for Word also counted toward that shared quota. Codex is already included in all ChatGPT plan tiers, and Codex Cloud is available to eligible Plus, all Pro tiers, Business, Enterprise, Healthcare, and Education accounts; usage depends on the model, run location, task complexity, context, reasoning, speed, and tools, with long-running tasks consuming significantly more. Codex limits are calculated across two windows, five-hour and weekly; Business Premium seats have no five-hour limit, while Standard seats allow roughly 15–150 messages per five hours on GPT-6.1 Sol or GPT-6 Sol.

    Same story · 10 reports

    #Product updates#OpenAI#English text is being prepared.

  10. Unite.AI
    AI score56

    Envision deploys 800V DC power supply system at Ulanqab AI data center

    Envision announced that its Galaxy AI data center in Ulanqab, Inner Mongolia, is now carrying actual computing loads with an 800V DC power supply system, which the company says is the first commercial application of this system in a large-scale AI data center. The system covers end-to-end power supply from medium-voltage AC input to server racks, integrating solid-state transformers, battery energy storage and DC voltage regulation equipment, with the Dubhe controller working alongside the Dubhe energy foundation model to adjust transformers and batteries in real time to follow GPU load fluctuations. Envision says the architecture can improve power conversion efficiency, reduce equipment footprint and cut cable and copper usage; the Galaxy campus is planned for 3GW capacity, with a full-load design supporting up to 1 million AI accelerators running in parallel.

    Same story · 2 reports

    #Industry news#English text is being prepared.#English text is being prepared.

  12. MediaNama
    AI score58

    IndiaMART moves content behind a login wall, restricting Google Search and AI agent access

    The Indian B2B classifieds platform IndiaMART has moved its content behind a login and restricted Google Search and AI agents from crawling its data. CEO Dinesh Agarwal said on the Q2 FY27 earnings call that the company is willing to give up the current 10-20% of standalone business inquiries to protect its data moat; standalone business inquiries for the quarter were 27 million, down 14% year over year. He noted that Google AI Overviews now covers 50% or more of searches, and that IndiaMART is open to licensing desensitized non-personal data, but only on the premise that it generates revenue; search and AI agent access remain restricted.

    #Industry news#Search#Agents

  13. MediaNama
    AI score69

    TRAI draft requires telecom operators to bear responsibility for the consequences of third-party AI systems; Anthropic prospectus says AI liability law is unsettled

    On October 30, the Telecom Regulatory Authority of India (TRAI) released draft guidelines on the responsible use of AI in telecom networks, requiring licensed telecom operators to bear responsibility for the consequences of all AI systems they use, whether the systems are self-developed, purchased externally, or a combination of both; the comment period runs until November 27. The draft requires operators to establish a board-approved AI policy, classify AI by high/medium/low risk, have all AI systems approved and registered by a designated official, test accuracy and robustness before launch, report AI incidents to TRAI in four tiers, and set up an AI governance and safety committee led by senior executives. In its SEC listing registration filed on October 26, Anthropic said the legal liability framework for the behavior of autonomous agents has not yet been determined, that a single-step error in a multi-step task can be amplified through subsequent steps, that the speed and scale of autonomous execution in enterprise deployments may exceed human review capacity, and that some traffic is subject to zero data retention (ZDR) terms, retaining only classifier scores and not prompts and outputs, which may fail to meet subpoena, regulatory review, or incident reporting requirements.

    #English text is being prepared.#Anthropic#Safety / alignment

  14. East Money
    AI score64

    The Hong Kong Securities and Futures Commission (SFC) has published its first AI adoption survey, setting out four key regulatory priorities

    On November 2, the Hong Kong Securities and Futures Commission (SFC) published its first "Survey on the Use and Preparedness of Artificial Intelligence among Large Institutions." The survey was conducted from July to August 2026 and covered 20 large financial institutions and their 78 affiliated licensed corporations in Hong Kong. The results show that 30% of the surveyed institutions have already applied agentic AI in their business operations, while another 25% are conducting pilots; 95% expect to accelerate deployment in the next 12 months, and 90% have already connected AI tools to internal systems. On this basis, the SFC put forward four key regulatory priorities: safeguarding investor interests; strengthening cybersecurity and operational resilience; implementing human accountability and human oversight in workflows with increasing autonomy; and building knowledge and capacity through exchange and collaboration.

    #English text is being prepared.#Agents#Safety / alignment

  15. Sina Tech
    AI score56

    Abu Dhabi and OpenAI sign four agreements covering government, energy, and investment

    Abu Dhabi-related institutions and OpenAI signed four agreements involving the government, energy, and investment sectors. On energy, ADNOC, G42's Inception42, and OpenAI signed an agreement, making ADNOC the first company in the energy sector to deploy OpenAI's frontier technology, with compute provided by inference infrastructure within the UAE. On government, Abu Dhabi and OpenAI launched the world's first AI-native Machine-Readable Government (MRG) initiative, with the first pilot already in production with ADAFSA, focusing on regulatory compliance assessment for food imports. On investment, Mubadala, Inception42, and OpenAI are collaborating to apply AI to investment analysis, opportunity identification, and risk assessment; Inception42 also signed a five-year agreement with OpenAI to promote the adoption of frontier AI in government, enterprise, and education.

    Same story · 3 reports

    #Industry news#OpenAI#English text is being prepared.

  16. Phoenix Technology
    AI score58

    South Korea unveils 2027 budget, with AI, semiconductors and data centers receiving 21.3 trillion won in support

    South Korean President Lee Jae-myung announced the 2027 government budget plan on November 2, proposing to raise total annual fiscal spending to 820.9 trillion won (about 614.5 billion US dollars), a year-on-year increase of 12.8%, setting a record high. Among this, projects related to semiconductors, physical artificial intelligence (Physical AI) and AI data centers are planned to receive 21.3 trillion won in support, with an additional 4.7 trillion won invested in developing independent artificial intelligence models. South Korea Customs Service data shows that total exports in the first 10 months of 2026 reached 930 billion US dollars, surpassing 900 billion US dollars for the first time.

    Same story · 3 reports

    #English text is being prepared.#English text is being prepared.#Industry news

  17. cnBeta
    AI score62

    OpenAI, Anthropic, Google, and Microsoft jointly draft a statement rejecting AI training

    OpenAI, Anthropic, Google, and Microsoft have jointly drafted an internet draft: websites only need to add User-Agent: __usage_ai-training and Disallow:/ to robots.txt to declare that all content on the site is prohibited from being used for AI training, and they can also set permissions or prohibitions by directory. The companies' support timelines differ: Microsoft says it will comply starting February 1, 2027, Google plans to support it in early 2027, OpenAI says it will implement it within the next few months, and Anthropic appears to have already begun supporting it. This statement only expresses a content usage preference and will not directly block crawler access; restricting crawling still requires traditional robots.txt rules.

    #English text is being prepared.#English text is being prepared.#OpenAI

  18. Phoenix Technology
    AI score78

    Anthropic says an unreleased Claude model found a Linux kernel vulnerability that lay dormant for 11 years, gaining Root in 53 minutes

    Anthropic disclosed that its internal unreleased Claude model, during systematic testing, discovered a serious Linux kernel vulnerability that had lain dormant for 11 years, CVE-2026-98432, with a CVSS score of 9.8, at the remote code execution level. The model also generated a working two-stage exploit: the first stage took about 48 minutes to defeat kernel address space randomization, and the second stage took 5 minutes to execute arbitrary commands with Root privileges, for a total of about 53 minutes. The upstream fix was completed on October 15, but as of November 1, Ubuntu 22.04 LTS, 24.04 LTS, 26.04 LTS, and 26.10 are still marked as affected, and the fix is still in development.

    Same story · 3 reports

    #Safety / alignment#Anthropic#English text is being prepared.

  19. cnBeta
    AI score84

    Claude discovers a severe Linux kernel vulnerability that lay dormant for 11 years and generates a working exploit

    Anthropic's Claude model discovered a use-after-free (UAF) flaw in the Linux kernel's IPv4 TCP handshake. The code was submitted in October 2015 and introduced with Linux 4.4; an attacker can trigger remote root privilege execution through network traffic without a password or user interaction. Claude also generated a working two-stage exploit: the first stage took about 48 minutes to defeat kernel address space layout randomization protections, and the second stage took only 5 minutes to execute arbitrary commands with root privileges. The whole process took about 53 minutes, and a failure does not cause the target to crash. The upstream fix was completed on October 15, covering 5.10.272, 6.1.190, 6.12.113, 6.18.56, 7.2.10, and 7.3-rc7, but as of November 1, Ubuntu 22.04 LTS, 24.04 LTS, 26.04 LTS, and 26.10 are still marked as affected, and the fix is still under development.

    #Safety / alignment#Anthropic

  20. 36Kr · AI
    AI score65

    Meta and Fourier Intelligence Partner to Build Humanoid Robot Prototype Max in Three Months

    According to foreign media reports, Meta partnered with Chinese robotics company Fourier Intelligence to develop a humanoid robot prototype named Max in about three months, and a second-generation product has already entered the R&D stage. Max is about 1.5 meters tall and weighs 48 kilograms, is equipped with NVIDIA's Jetson Thor chip, has five-fingered dexterous hands and a display screen that can simulate human facial expressions, and needs to recognize human voices and gestures and maintain eye contact. Li-Chen Miller, Meta's Vice President of Robotics Product, was stationed in Shanghai for nearly two months to help advance the effort, and Meta is also recruiting for supplier development and mass-production preparation roles in Shanghai and Shenzhen.

    Same story · 6 reports

    #Industry news#Embodied AI#Meta