NAll news

Thursday208 items

  1. Smart City Industry Analysis
    AI score55

    Jiangsu Issues Action Plan for Building High-Quality Datasets for Physical AI and World Models (2026–2028)

    The Jiangsu Provincial Data Bureau and three other departments issued the "Jiangsu Province Action Plan for Building High-Quality Datasets for Physical AI and World Models (2026–2028)," proposing that by the end of 2028, high-quality physical AI datasets covering more than 100 key scenarios be built, along with more than 10 high-quality datasets oriented toward frontier directions of world models. The plan lists 5 major items and 14 key tasks, covering directions such as building real-machine interaction datasets, expanding simulation and synthetic data, and tackling frontier datasets for scientific research. In the 2026 project list issued at the same time, 39 projects were selected as key projects, and 79 projects from 12 prefecture-level cities were selected as construction and cultivation projects, covering 15 tracks.

    #English text is being prepared.#English text is being prepared.#Embodied AI

  2. The Paper
    AI score87

    AI-Assisted Discovery of Critical PHP Vulnerability Prompts Emergency Shutdown of Government Websites in Multiple European Countries

    Cybersecurity firm Calif, in collaboration with OpenAI, used AI models to discover a critical vulnerability in the PHP programming language; Wiz also independently discovered a related vulnerability, and the PHP development team released a security update on October 21. The European Union Agency for Cybersecurity said the vulnerability could allow attackers to remotely execute malicious code without authentication, and some government websites and public service platforms in the Netherlands, Finland, France, and Germany were temporarily taken offline. Censys counted approximately 2.58 million internet-accessible PHP hosts, of which about 1.576 million reported versions earlier than the fixed version; Tencent Cloud, Alibaba Cloud, and QiAnXin CERT also issued risk advisories in parallel.

    Same story · 8 reports

    #Safety / alignment#OpenAI#Google

  3. Latent Space
    AI score79

    OpenAI Patch the Planet used GPT-5.6 Sol to discover the PHP2Shell vulnerability

    OpenAI says researchers in its Patch the Planet project used GPT-5.6 Sol to discover PHP2Shell (CVE-2026-109618 and CVE-2026-109619), a core PHP vulnerability affecting about 70% of websites, and coordinated the fix with the PHP maintainers and others over roughly two months. Wiz, for its part, says it independently discovered the same set of vulnerabilities using Gemini 4 Argon, and Cloudflare deployed free WAF mitigation rules before the patch release, describing them as only a temporary measure. The original article also notes that key evidence is still missing, including the specific affected and fixed versions, any pre-patch leak of the content, and actual exploitation in the wild.

    Same story · 6 reports

    #Safety / alignment#OpenAI#Google

  4. CTech
    AI score88

    AI discovers a PHP vulnerability over a decade old; government websites in multiple European countries urgently suspend services

    Two cybersecurity companies used AI to discover a critical vulnerability in PHP that had been hidden for over a decade, and government agencies and organizations in multiple European countries suspended online services on Wednesday. Censys identified about 2.6 million hosts using PHP, of which about 1.5 million run versions with no fix available; Wiz said it completed the discovery using Wiz Atlas with Gemini 4 Argon and GPT-5.6 Sol, while Calif said it discovered the vulnerability on August 1 with the help of a large language model. Amazon security chief CJ Moses said its honeypot system had detected exploitation attempts four days before public disclosure.

    #Safety / alignment#Agents#OpenAI

  5. Wallstreetcn
    AI score69

    SoftBank's stock price halves from its June high, AI financing risks spark market concerns

    SoftBank Group's stock price fell 6.9% in a single day on Thursday, accumulating a 50% decline since its historical high on June 2, with market capitalization shrinking to about 24.7 trillion yen (approximately $157 billion). The trigger for this selloff was cloud service provider CoreWeave's bond prices falling into distressed territory, sparking market concerns about financing risks in the AI industry; SoftBank's CDS spread rose on Tuesday to its highest level in nearly four years. SoftBank's investment in OpenAI has reached nearly $65 billion, and it had just raised $11.1 billion in September through the issuance of corporate junk bonds.

    #Industry news#OpenAI

  6. CTech
    AI score61

    Meta advances AI agent strategy as Q3 free cash flow turns negative at $2.12 billion

    Meta is making AI agents its new growth engine while ramping up infrastructure spending and seeing its cash flow deteriorate. The company reported Q3 revenue of $64.6 billion, above analysts' expectations of $63.37 billion, and raised the lower bound of this year's capital expenditure from $130 billion to $135 billion; free cash flow for the quarter was negative $2.12 billion, compared with positive $10.63 billion a year earlier. The consumer-facing agent app Muse surpassed 10 million users six weeks after launching in the US and Canada, and Meta plans to issue non-voting Class C common stock in January 2027, expecting its AI business to begin contributing incremental revenue starting in 2027.

    #Industry news#Agents#OpenAI

  7. Phoenix Technology
    AI score82

    DeepSeek reportedly completes about 100 billion yuan financing, with CATL and Tencent participating

    According to Bloomberg, DeepSeek raised about 100 billion yuan (about $15 billion) in its latest funding round, setting the record for the largest financing in China's AI industry, with CATL and Tencent among the largest investors in this round. People familiar with the matter said that during its discussions with investors, DeepSeek raised its financing target at least twice, from an initial 50 billion yuan to 80 billion yuan, and ultimately to 100 billion yuan. Bloomberg said the funds will be used to advance the construction of AI data centers, develop more open-source models for the Chinese and overseas markets, and provide a funding base for DeepSeek's preparations for an IPO; Reuters previously reported that it had hired CITIC Securities to prepare for an IPO on the Shanghai Stock Exchange's STAR Market.

    Same story · 14 reports

    #Industry news#DeepSeek

  8. IT Home · Artificial Intelligence
    AI score63

    SoftBank's stock price has nearly halved from its June high, and the money-burning model of AI investment is being questioned by the market

    According to Bloomberg, SoftBank's stock price fell as much as 6.7% during intraday trading, bringing its cumulative decline from its all-time high on June 2 to 50%, and it has fallen more than 30% this month. Market concerns have intensified because CoreWeave's bond prices have fallen to distressed-debt levels, and the quote for SoftBank's credit default swaps, used to hedge against the risk of debt default, has risen to a nearly four-year high. SoftBank has invested nearly $65 billion in OpenAI and in September issued $11.1 billion in junk bonds to support Masayoshi Son's AI investment plan.

    #Industry news#OpenAI

  9. 36Kr · AI
    AI score64

    Meta's free cash flow turned negative at $2.12 billion in Q3 2026, capital expenditure guidance raised to $135–145 billion

    Meta reported Q3 2026 revenue of $64.601 billion, up 26% year-over-year, but capital expenditure rose to $33.942 billion and free cash flow fell to negative $2.12 billion, compared with positive $10.625 billion in the same period last year. The company raised the lower bound of its full-year 2026 capital expenditure guidance by $5 billion to $135–145 billion, and Chief Financial Officer Susan Li said investment will continue to increase in 2027. The earnings report also disclosed a plan to establish non-voting Class C common stock for purposes such as employee equity incentives and potential acquisitions.

    Same story · 9 reports

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

  10. 36Kr · AI
    AI score75

    Meta's Q3 revenue was $64.601 billion; Muse surpassed 10 million setups within six weeks of launch

    Meta's revenue for the third quarter of fiscal year 2026 was $64.601 billion, up 26% year over year and a new single-quarter high, but a $9.89 billion litigation settlement expense caused operating profit to fall 37% year over year, and free cash flow was negative $2.120 billion. On the earnings call, Zuckerberg disclosed that about six weeks after the AI product Muse launched, more than 10 million people in the U.S. and Canada had completed setup; shopping is the most popular use, with an average purchase amount exceeding $200, and transaction volume in the first week after launch exceeded expectations, requiring Meta to ask Stripe to raise its credit limit. He said Muse will move toward commercialization earlier than previous new products, with the basic version free and subscriptions at $20 and $100 per month, and plans to bring it to smart glasses in November and deliver the dedicated device Charm in December.

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

  11. 36Kr · AI
    AI score56

    Australian startup Apate.ai closes US$8.15 million seed round, using 350,000 AI bots to fight scams

    Australian company Apate.ai has closed a US$8.15 million seed round led by Lobby Capital, with participation from OIF Ventures, Investible, Concept Ventures and Baobab Ventures, bringing cumulative funding to nearly US$10 million. The company uses large-model-driven AI to impersonate ordinary people and keep scammers occupied; it is reported to have completed about 2.5 million conversations and to have roughly 350,000 AI bots, with customers including Commonwealth Bank of Australia and TPG Telecom. Cao Xiao of Shanghai University of Finance and Economics argues that intelligence obtained through such proactive engagement should be subject to a tiered mechanism and multi-source cross-verification, and should not be used directly as grounds for blocking.

    Same story · 2 reports

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

  12. TMTPost · AGI
    AI score62

    Meta's earnings report discloses Muse usage data; Zuckerberg says commercialization will be pushed forward earlier

    On Meta's Q3 fiscal 2026 earnings call, the company disclosed that about six weeks after Muse launched, more than 10 million people in the U.S. and Canada had completed setup, with shopping being the most popular use case and an average purchase amount exceeding $200. Zuckerberg said Muse will advance commercialization earlier than the company's previous new products because both the costs and the business opportunity are large enough. Meta's Q3 revenue reached a record single-quarter high of $64.601 billion, but due to $9.89 billion in litigation settlement expenses, operating profit fell 37% year over year, and free cash flow was negative $2.120 billion.

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

  13. MediaNama
    AI score53

    ixigo Responds to AI Agent Disruption: Building Its Own Agent Layer and Serving as a Third-Party Agent Supplier

    In its Q2 FY27 earnings report, ixigo said it will simultaneously serve as a supply and fulfillment partner for third-party AI agents, and provide its own agent experience through ixigo NEXT, a voice-first layer built on its in-house small language models. Management said that when users shift to external agents, the app loses engagement, context, and even its distribution channel, but believes the Indian market is harder for general-purpose agents to penetrate: trains and buses contribute more than 70% of its contribution margin, and 94% of transactions come from travelers in tier-2, tier-3, and tier-4 cities. Its AI assistant TARA has resolved 94% of chat queries and handled 82% of customer service calls end to end; in the same period, technology spending rose 46% year over year to INR 14.3 crore, while adjusted EBITDA fell 17% to INR 23.6 crore.

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

  14. BBC Technology
    AI score86

    AI discovers years-old PHP vulnerabilities, institutions in multiple countries worldwide urgently take websites offline

    Researchers from security companies Calif and Wiz, using OpenAI technology, discovered vulnerabilities in PHP that could allow attackers to take over servers and steal data without authentication; Wiz called it the most severe vulnerability detected by AI. PHP developers have released patches and urged all users to update immediately; the US NSA advised organizations unable to patch immediately to take systems offline, and the Finnish government, Edinburgh Council, the BBC, the Dutch cybersecurity agency, and multiple Scottish public institutions have taken their websites offline. OpenAI called the discovery one of the most impactful security issues recently and released a program to check whether PHP platforms are secure.

    Same story · 8 reports

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

  15. The Guardian Technology
    AI score63

    Nvidia, Microsoft and others warn UK has one year to defuse data center protest risk

    AI infrastructure companies including Nvidia, Microsoft and Cisco have submitted recommendations to the UK government, saying the UK has about 12 months to establish a "compact" that benefits communities, or it may face protests similar to those in the US that threaten data center construction. The SAIL working group recommends letting councils that approve data centers retain more tax revenue and establishing a local benefit framework. Organizations such as PauseAI UK and Foxglove criticized this approach as buying off communities with benefits, and noted that 15 community groups in Scotland are already opposing related projects.

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

  16. Ayi AI Notes
    AI score55

    Musk Unveils Neuralink Visual Implant Blindsight

    Musk has unveiled Neuralink's visual implant Blindsight, which bypasses the eyeballs and writes images directly into the brain's visual cortex; the first-generation product offers low-resolution black-and-white pixels. The author believes its endpoint is to give humans multispectral vision including infrared, ultraviolet, and even radar waves, but clinical approval progress, whether the brain can decode heterogeneous data streams, and the electrode scarring and rejection associated with long-term implantation all await data from the first batch of subjects.

    English text is being prepared.

    Same story · 2 reports

    #Industry news#Embodied AI

  17. cnBeta
    AI score74

    Report: DeepSeek raises 100 billion yuan, with Tencent and CATL participating

    People familiar with the matter said that during negotiations with investors, DeepSeek raised its funding target at least twice, from an initial 50 billion yuan to 80 billion yuan, with the final funding amount reaching 100 billion yuan. CATL and Tencent were among the largest investors in this round. This round of funding paves the way for its future listing, but DeepSeek's financial strength still lags behind Anthropic and OpenAI, which have each raised more than $100 billion. DeepSeek has increased investment in AI data centers and released V4.1 Flash in September this year, triggering declines in the share prices of MiniMax and Zhipu.

    Same story · 14 reports

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

  18. cnBeta
    AI score53

    Trump demands AI data centers supply surplus power to communities

    Trump warned that AI data centers must not only generate their own power but also must supply surplus electricity to local communities for free or sell it to them, and called for construction to proceed after reaching agreements with residents. The report notes that communities across the United States are protesting data center projects over issues such as rising electricity prices, noise, and water shortages. Hazel Township, Pennsylvania, once proposed compensating residents who agreed to data center construction with $10,000, but many residents still opposed it. The Trump administration had previously required such companies to generate their own power, and recently passed legislation requiring data centers to bear the costs of grid upgrades.

    Same story · 5 reports

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

  19. cnBeta
    AI score76

    Sources say Anthropic could file for IPO as early as Friday

    Anthropic confidentially submitted a draft IPO registration statement to the SEC in June, and is reported to make the S-1 filing public as early as this Friday, clearing procedural hurdles for a November listing. According to The Wall Street Journal, the company posted revenue of $16.3 billion and a net loss of more than $80 billion in the first half of 2026; the Financial Times said its third-quarter revenue rose to $16.8 billion, up from $11.5 billion in the second quarter, and it achieved profitability on an adjusted basis. Reuters previously reported that Wall Street expects its valuation could reach $2 trillion, with fundraising of up to $100 billion.

    #Industry news#Anthropic#OpenAI

  20. cnBeta
    AI score85

    OpenAI and PHP maintainers fix critical PHP2Shell vulnerability

    OpenAI and PHP maintainers coordinated to fix two critical PHP engine vulnerabilities collectively called PHP2Shell, both with a CVSS score of 9.2, which can lead to unauthenticated remote code execution. The vulnerabilities were discovered by a Calif researcher with the help of GPT-5.6 Sol and independently reproduced by Wiz; the fix changed PHP's memory management mechanism, and PHP 8.5.12, 8.4.27, 8.3.36, and 8.2.35 have been released. The unified release originally scheduled for the evening of October 22 was moved up by one day because the vulnerability IDs and an attack demonstration leaked early, and Cloudflare had already observed early attacks before the patch went live.

    Same story · 6 reports

    #Safety / alignment#Agents#OpenAI