要闻提示 NEWS REMIND 1.为何停售初代小米SU7？雷军：不想背刺老车主 2.华为渠道服突发大规模封号！成年玩家无辜躺枪，客服回应：请重新登陆 3. AI大厂月薪3w疯抢文科生！网友：拯救了文科就业率 4.周鸿祎：决定启动360安全龙虾全国巡装计划 5.传地平线芯片研发负责人陈鹏即将离职 6.阿里正推进一项内部计划：向员工提供Token额度，鼓励使用AI工具工作 7.三星将停售首款三折叠手机！上市仅3个多月，曾在中国市场遭抢购 8.遣散费近40亿！PC巨头裁员11000人 今日头条 HEADLINE NEWS 为何停售初代小米SU7？雷军：不想背刺老车主 3月17日消息，小米创始人、董事长兼 CEO 雷军在其官方微博账号发布《新一代 SU7 雷军答网友问（第 二集）》。在回应"为什么这么早停售第一代 SU7，不担心销量吗？"时，雷军表示，"（停售）确实对我 们前三个月的销售产生了影响，但我们不想背刺老用户。"除了停售相关问题，雷军在视频中还回应了诸 多大众关注的话题。 谈及小米汽车过去两年的发展历程，他用"跌宕起伏、冰火两重天"来概括，坦言小米造车所取得的成绩远 超预期，但同时也收到了大量批评与质疑 ...

完整早报音频，请点击标题下方小耳机收听 【马斯克点赞Kimi新架构，17岁高中生作者一战成名】 AI领域的创新先锋月之暗面（Kimi）团队近日发布了一项 突破性的研究，提出名为"Attention Residuals"的神经网络架构改进技术，成功将Transformer中的核心"注意力 机制"理念创新性地应用于网络深度维度，显著提升了大模型的训练效率。该研究引发了全球AI社区的高度关注，包 括埃隆·马斯克在内的多位科技领袖均在社交网络上表达了赞叹。 这项重磅研究由三位研究者共同担任第一作者，其中最引人注目的是年仅17岁的高中生研究员陈广宇。另外两位共 同一作分别是Kimi核心成员、著名旋转位置编码RoPE的提出者苏剑林，以及Kimi Linear架构的第一作者张宇。 （创业邦） 【蚂蚁集团董事长井贤栋向母校上海交大捐赠 1.3 亿】 3 月 17 日，上海交通大学建校 130 周年之际，井贤栋捐 赠签约暨校董聘任仪式在闵行校区文博楼举行。蚂蚁集团董事长、 1994 届本科校友井贤栋与同为 1994 届本科校 友的夫人，共同向母校捐赠价值 1.3 亿元的现金和蚂蚁集团股份，用于支持上海交大" AI 未来基石基金 ...

3 月 16 日晚间，360 就旗下 AI 产品"安全龙虾"私钥泄露事件给出官方回应，确认系产品发布环节操作失误， 目前已完成应急处置。 SSL 私钥是 HTTPS 通信的核心凭证， 把私钥明文放进公开安装包，等于把家门钥匙直接丢在大街上 ，理论 上可被用于伪造服务器、劫持流量、窃取通信数据，对安全产品而言属于严重疏漏。 360 在回应中表示，此次问题是内部测试证书被误打包进公开发布包，并非主动泄露。公司已第一时间吊销涉 事证书，证书现已完全失效，从技术上阻断被恶意利用的可能。 事件暴露出 AI 产品快速迭代中的安全短板。主打安全的厂商出现基础安全失误，也引发行业对发布流程、权 限管控、上线审计的讨论。 （参考： X 、一财，本文经由 AI 优化） 3 月 14 日，360 正式推出主打安全的 OpenClaw 智能体产品"安全龙虾"，主打快速部署与安全防护。 仅两天后，安全社区研究人员在公开安装包中，发现明文存储的 *.myclaw.360.cn 泛域名 SSL 证书与私 钥，事件迅速发酵。 「安全吗？ 20 年安全老本行，你觉得呢？」 ...

Group 1 - ByteDance has suspended the global release of its video generation model Seedance 2.0 due to copyright disputes with major Hollywood studios, including Disney and Paramount's Skydance [2] - Kimi's valuation has surged to $18 billion, marking a fourfold increase in just three months, as the company is currently raising $1 billion in a new funding round [2] - Founder Yu Donglai of Pang Donglai clarified the asset distribution of 4 billion yuan, stating that his personal stake is around 5%, which is approximately 200 million yuan [3] Group 2 - ByteDance responded to rumors of layoffs at its Wuhan R&D center, stating that only 50 employees will be relocated, and the center will continue to operate with over 2,000 employees [4] - 360 Group launched the "360 Security Lobster" intelligent application client, significantly reducing installation time from 6 hours to 10 minutes [8] - Honda is expected to pay suppliers up to 106 billion USD due to the cancellation of three electric vehicle projects as part of a strategic reassessment [8] Group 3 - Giant Quasar Electronics completed a financing round of several million yuan to enhance its core technology and expand its market presence in high-end electronic testing [9] - Micro-LED manufacturer Sitan Technology raised over 200 million yuan in Series B funding to accelerate commercialization in automotive and AR display sectors [9] - Xi'an Zhongke Optoelectronics completed nearly 400 million yuan in Series A financing, focusing on intelligent manufacturing breakthroughs in aerospace and shipbuilding [9] Group 4 - Chery's iCAR V27, a new range-extended SUV, was launched with a price range of 169,800 to 196,800 yuan, featuring a comprehensive range of over 1200 km [10] - Qianwen AI glasses will introduce new features such as AI simultaneous translation and "AI handling" by the end of the month, expanding their application in daily services [12] - The overall passenger volume for civil aviation during the 2026 Spring Festival reached 94.39 million, a historical high, with a daily average of 2.36 million passengers, reflecting a 4.7% year-on-year increase [14]

据陕西都市快报等媒体消息，3月14日，西安交通大学在兴庆校区科学馆101举行"百卅交大·伴爱同行"捐赠仪式。会上公布，360集团创始 人、董事长，西安交大校友周鸿祎，向母校捐赠1亿元。 公开报道显示，周鸿祎1988年至1992年就读于西安交大计算机软件专业，获学士学位，1992年至1995年就读于西安交大系统工程专业， 获硕士学位。 ...

来源丨凤凰网财经 作者丨王迪 3月12日，360集团创始人周鸿祎针对近期刷屏AI圈的OpenClaw（俗称"龙虾"）智能体，分享了他对安全争议、爆火逻辑、产业变革与商业化等核 心话题的看法。 对于AI及智能体快速发展可能导致人类变懒、能力退化的担忧，周鸿祎在与凤凰网财经的交流中给出回应：真正用好智能体并非让人"躺平"，反而需 要使用者更深度地参与、操练和指挥，这将催生人类新的核心能力。 在周鸿祎看来，AI智能体并非取代人类的"自动化黑箱"，而是需要人类深度介入、共同进化的"数字同事"或"数字员工"。未来的竞争差异，不在于是 否拥有智能体，而在于谁更善于"操练"和"指挥"智能体。人类的核心角色将从"操作者"转向"教练员"与"指挥官"，这要求更高的沟通、判断、管理 和战略能力，而非导致退化。 此外，在交流现场，周鸿祎还指出，龙虾不是工具，是有自我进化能力的硅基物种，是智能体的终极形态；未来不是卖软件，而是卖龙虾、卖数字劳 动力；不用智能体的人，终将被会用智能体的人淘汰；智能体消耗的算力，是聊天机器人的100倍、1000倍，算力需求将迎来爆发。 360集团创始人周鸿祎 以下为周鸿祎与媒体交流的核心观点： 01 回应 ...

Group 1 - Honda announced a significant loss of 1 trillion yen due to the withdrawal of its electrification strategy, with expected operating losses of 270 billion to 570 billion yen for FY2025 [2] - Tencent's SkillHub platform faced accusations of plagiarism from OpenClaw's creator, but Tencent clarified that it only serves as a local mirror for Chinese users and has contributed to the original project's bandwidth [2] - The launch of the Chasing Car series at AWE2026 is positioned against Xiaomi's SU7 Ultra, with a price range of 600,000 to 700,000 yuan and impressive performance metrics [7] Group 2 - Li Auto reported a revenue of 112.3 billion yuan and a net profit of 1.1 billion yuan for the fiscal year, marking three consecutive years of profitability [9] - ByteDance welcomed Yu Bowen, a former Alibaba executive, to lead its visual model and multimodal interaction team, indicating a strategic move in AI development [9] - Nvidia plans to invest 26 billion USD (approximately 178.8 billion yuan) over the next five years to develop open-source AI models, significantly surpassing OpenAI's investment in GPT-4 [21] Group 3 - BYD is conducting large-scale recruitment, with over 90,000 employees, making it the largest employer among A-share listed companies [13] - Apple celebrated its 50th anniversary, with CEO Tim Cook expressing gratitude to users for their role in shaping the company's innovations [14][15] - Atlassian announced a 10% workforce reduction, affecting around 1,600 employees, to adapt to the rise of AI and strengthen its financial position [21]

Investment Rating - The industry investment rating is "Positive" and maintained [8] Core Insights - Recently, multiple internet and model manufacturers have followed OpenClaw, launching various ready-to-use versions of "lobster," and related policies have been introduced in several regions, leading to a surge in the "lobster farming" concept [2][5] - The current trend driven by OpenClaw indicates a significant shift in AI deployment capabilities, moving towards executable, deployable, and privatized solutions, with competition focusing on execution capabilities and commercialization [11] - The demand for AI agents is expected to see a turning point, with the introduction of free trials, low-cost subscriptions, and lightweight deployments lowering usage barriers, while OpenClaw transitions AI from mere conversation to task execution [11] - The model market is expanding, and the token industry chain is becoming a new focal point, with anticipated growth in token revenue and demand for computing power as user installations increase [11] Summary by Sections Event Description - Multiple internet and model manufacturers are launching various versions of OpenClaw, with policies emerging in several regions to support this trend [5] Market Dynamics - Major companies are competing on deployment thresholds, usage costs, and service delivery, with at least 14 manufacturers currently involved in related products [11] - Policies in regions like Jiangsu and Guangdong are encouraging the development of industrial models using OpenClaw, with financial incentives for companies [11] Future Outlook - The report suggests that the AI industry is entering a demand era, with a focus on high-value applications in sectors like taxation and industry, as well as the potential for supply relief in domestic computing power [11]

Core Viewpoint - OpenClaw, an AI agent, is gaining popularity and prompting government support for its deployment, but it also presents significant security challenges due to its autonomous capabilities and system resource access [1][3]. Group 1: OpenClaw Overview - OpenClaw integrates communication software with large language models to autonomously perform complex tasks like file management and data processing [1]. - The tool's ability to call system resources and execute commands autonomously raises new security concerns [1]. Group 2: Security Risks - The Ministry of Industry and Information Technology has issued security warnings, indicating that even with updates, risks remain due to the agent's decision-making capabilities and complex skill sources [3]. - High system permissions required for OpenClaw can lead to severe consequences, such as data breaches or system control loss if security measures are inadequate [3][4]. Group 3: Expert Insights - Industry experts, including Zhou Hongyi, emphasize that while OpenClaw has innovative potential, it is still in its early stages, with high usage barriers and stability issues [3][4]. - Zhou compares AI agents to "interns" that require training and strict rules, highlighting the need for caution in their deployment [4]. Group 4: Security Guidelines - 360 Group has released a security deployment guide for OpenClaw, outlining various risks such as API key leaks and injection attacks [4][6]. - The guide recommends a "controlled first, then efficiency" approach, suggesting the use of containerization and minimal privilege strategies for individual developers and small teams [6]. Group 5: Implementation Strategies - For enterprise-level applications, a zero-trust security architecture is advised, including security gateways and fine-grained permission management [6][7]. - Integrating operational logs into security platforms can help identify and mitigate high-risk behaviors in real-time [7]. Group 6: Industry Implications - OpenClaw and similar AI agents are expected to transform industries like cloud computing did, but security capabilities must be developed concurrently to avoid high-risk costs in large-scale applications [7].

Core Viewpoint - The government work report emphasizes the development of a "new intelligent economy" and the integration of technology and industry, highlighting the importance of artificial intelligence (AI) and its application across various sectors [4][5][6]. Group 1: Intelligent Economy - The report introduces the concept of "new intelligent economy," calling for the promotion of new intelligent terminals and AI applications in key industries [4][5]. - Industry leaders express strong support for the government's focus on AI, indicating a shift from AI as a mere technology to a driver of value creation across sectors [5][6]. - The report outlines the need for deep integration of technology innovation and industrial innovation, marking a significant trend in the economy [6]. Group 2: Technological Infrastructure - The report mentions the implementation of large-scale intelligent computing clusters and the need for coordinated power supply, suggesting a focus on enhancing the efficiency of energy and computing resources [7]. - Recommendations include establishing a unified national electricity market to support intelligent computing centers and reduce energy costs for AI development [7]. Group 3: Future Industries - The report emphasizes the cultivation of future industries such as quantum technology, embodied intelligence, and 6G, indicating a strategic focus on advancing new productive forces [12][15]. - Industry representatives highlight the importance of developing quantum communication and embodied intelligence, with a call for government support in overcoming common challenges in these fields [12][15]. Group 4: International Logistics and Trade - The report stresses the need to strengthen international logistics systems and expand digital trade, which is seen as beneficial for companies engaged in global supply chain services [16][17]. - Companies are adapting to a shift from "stocking mode" to "branding mode" in cross-border logistics, necessitating higher standards for logistics stability and compliance [16]. Group 5: AI and Data Security - The report raises concerns about data security and user authorization in AI applications, highlighting the need for regulatory frameworks to address potential issues arising from data usage [9][10]. - Experts emphasize the importance of user rights and data protection in the deployment of AI agents, advocating for clear guidelines on data handling and user consent [10].