VOL.2026.07.28 · 30 篇报道 · AI 日报
AI 日报 — 2026-07-28
星期二 · 30 篇报道 · 约 15 分钟读完
今日的AI领域呈现出显著的两面性:一方面是能力上的惊人进步,另一方面则是日益增长的安全和控制担忧。尽管代理式AI在密码分析和科学计算等领域展现出前所未有的强大能力,但Hugging Face数据泄露事件以及Anthropic成功攻击加密系统,都凸显了关键的漏洞。这种紧张关系强调了对强大安全措施和协调治理的迫切需求,正如AI领导者们呼吁放缓发展速度,以确保在快速创新中的安全和稳定。
- 01模型与开源OpenAI和Anthropic对“Pacing the Frontier”倡议的支持,以及1100多名AI员工的联名信,表明行业领导者们对前沿AI发展采取更谨慎态度的共识日益增强,认识到在快速创新中治理的必要性。6
- 02Agent 与工具Anthropic的Claude Mythos Preview成功破解了AES的弱化版本,并对HAWK加密系统进行了改进攻击,展示了代理式AI在识别和利用安全漏洞方面的先进能力。7
- 03应用落地Anthropic公布了针对HAWK-256的实际密钥恢复攻击,源于AI代码创建,这突显了AI在加密系统中既能创建也能利用漏洞的潜力,强调了关键的安全问题。2
- 04融资&商业韩国芯片制造商三星电子和SK海力士股价分别暴跌超过11%和12%,影响了更广泛的市场,这表明投资者对全球经济变化背景下半导体行业前景的担忧。4
- 05政策&风险OpenAI、Anthropic和谷歌等主要AI公司的领导者签署了一份声明,敦促美国政府考虑放缓前沿AI发展或加速全球协调治理,这反映了对负责任AI进展的共同担忧。3
- 06行业动态包括John Schulman和Jakub Pachocki在内的1100多名AI公司员工签署了一封信,要求美国政府“放缓”AI发展,这表明行业内部普遍存在对更受控和深思熟虑的进展的呼声。8
01模型与开源6 篇
- #2
- #4Our position on open-weights models0 个来源 · 热度 53
- #16
- #22
- #27
- #29Satya Nadella says companies that trust one AI for everything may not survive
Microsoft CEO Satya Nadella warned that companies relying solely on proprietary AI labs for all their AI needs might not survive. He emphasized the importance of retaining metadata from AI model usage to train proprietary weights or open models. Nadella suggests businesses should maintain control over their usage data, enabling them to eventually develop their own AI models rather than depending entirely on external, single-source AI solutions.
0 个来源 · 热度 26
02Agent 与工具7 篇
- #1OpenAI just open-sourced Codex Security
OpenAI has open-sourced @openai/codex-security, a CLI and TypeScript SDK designed to identify, validate, and remediate security vulnerabilities in code. This tool allows users to scan repositories, review changes, track findings, and integrate security checks into their CI pipelines. It requires Node.js 22+, Python 3.10+, and access to Codex Security, with detailed documentation available for setup and usage.
0 个来源 · 热度 60 - #5Scientific computing in the age of agentic AI
Scientific computing is crucial for modern research, but its software often lags due to origins in academic teams with limited engineering. This leads to slow, fragile workflows that hinder discovery. Agentic AI, like GPT-5.5, can accelerate this by modernizing tools; for example, GPT-5.5 improved cyvcf2's build and packaging, making it easier to install, test, and release. As coding agents advance, researchers can focus more on discovery and less on maintaining analysis pipelines.
1 个来源 · 热度 39 - #8Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
Hugging Face detailed a July 2026 agent intrusion, reconstructing approximately 17,600 attacker actions between July 9 and July 13, 2026. The forensic analysis, aided by the open-source model GLM 5.2, revealed two initial-access vectors. One vector involved the agent committing a dataset with a configuration pointing to HDF5 files, allowing it to read raw bytes from local filesystem paths via the HF API, effectively disclosing files without executing code.
1 个来源 · 热度 36 - #9Gemini API Managed Agents: 3.6 Flash, hooks, and more
The Gemini API is expanding Managed Agents, introducing features like background tasks and remote MCP. Developers can now define custom domain functions, such as `get_weather`, and invoke agents with both built-in code execution and these custom functions. The `GoogleGenAI` client facilitates interactions, allowing for tasks like checking weather, executing Python scripts, and handling custom function execution cleanly. The `interactions.create` method supports specifying an agent, input, environment, and tools, including `code_execution` and custom tools.
1 个来源 · 热度 33追踪这条信号 - #17Anthropic says Claude Mythos Preview was able to break a weaker version of AES and orchestrated another improved attack against the cryptographic system HAWK (Dustin Volz/New York Times)
Anthropic's Claude Mythos Preview reportedly broke a weaker version of AES and orchestrated an improved attack against the HAWK cryptographic system. This AI discovered new attacks during testing against weakened cryptographic algorithms. These algorithms are crucial for protecting online financial transactions and private communications, highlighting the potential for AI to identify vulnerabilities in security systems.
0 个来源 · 热度 27追踪这条信号 - #20
- #25
03应用落地2 篇
- #3Anthropic publishes a practical key-recovery attack on HAWK-256
Anthropic has published details of a practical key-recovery attack targeting HAWK-256. This development, related to AI CODE CREATION, highlights potential vulnerabilities in cryptographic systems. The information was shared within the developer community, indicating its relevance for those involved in enterprise-grade security and premium support, emphasizing the ongoing need for robust cryptographic solutions.
0 个来源 · 热度 54追踪这条信号 - #6
04融资&商业4 篇
- #18NVIDIA Cosmos-H-Dreams: Bringing Real-Time Generative Simulation to Surgical Robotics
NVIDIA Cosmos-H-Dreams introduces real-time generative simulation to surgical robotics, addressing the challenges of evaluating and training vision-language-action policies. The Cosmos-H-Surgical-Simulator, built on NVIDIA Cosmos-Predict2.5-2B and Open-H-Embodiment, generates video consequences of robot trajectories, enabling offline policy evaluation and synthetic data generation. This innovation allows for practice, exploration, and data generation without the costs and risks associated with physical robotic platforms, making surgical simulation more accessible and efficient.
1 个来源 · 热度 27追踪这条信号 - #21
- #28Fish Audio raises $52M seed to build AI voice models for creators and enterprises
Fish Audio has secured $52 million in seed funding to develop AI voice models for creators and enterprises. The company originated from a project by former Nvidia researcher Shijia Liao, who open-sourced a voice-generation model on a single GPU due to the lack of expressive synthetic voices. This open-source project, Fish Speech, has garnered over 31,000 stars on GitHub and is utilized by indie developers, video game designers, and other creators, addressing the market's need for more expressive and steerable AI voices.
0 个来源 · 热度 26 - #30At market open Korean chipmakers' stocks plunged, with Samsung Electronics dropping 11%+, SK Hynix down 12%+, and the Korea Composite Stock Price Index down 9%+ (Youkyung Lee/Bloomberg)
Korean chipmakers' stocks experienced a significant plunge at market open, with Samsung Electronics dropping over 11% and SK Hynix falling more than 12%. This downturn also impacted the Korea Composite Stock Price Index, which saw a decline of over 9%. The plunge was attributed to worsening investor sentiment regarding the artificial intelligence boom, particularly affecting the nation's memory chipmaking powerhouses.
0 个来源 · 热度 26
05政策&风险3 篇
- #12The OlmoEarth Platform: Geospatial inference at planetary scale
The OlmoEarth Platform, detailed at allenai.org/olmoearth, enables geospatial inference at a planetary scale. Its independent partitions allow stages to run across thousands of compute instances simultaneously. This was demonstrated by generating a North American wildfire-risk map, utilizing 19,600 CPUs and 994 GPUs in parallel, achieving a 155x speedup by reducing 4,737 hours of serial compute to 30.5 hours. The platform also aims to lower barriers to geospatial model use through agentic tools and interfaces.
0 个来源 · 热度 29追踪这条信号 - #15AI leaders sign a statement asking the government to do something about automated AI
Leaders from OpenAI, Anthropic, Google, Meta, Microsoft, and other AI labs have signed a statement urging the US government to consider a slowdown in frontier AI development or accelerate global coordinated governance efforts. Notable signatories include OpenAI's Mark Chen, Jakub Pachocki, John Schulman, and Wojciech Zaremba, along with Anthropic's Jack Clark, Chris Olah, Ben Mann, and Jared Kaplan. Former OpenAI executives Josh Achiam and Jan Leike also signed, advocating for government intervention regarding automated AI.
0 个来源 · 热度 27 - #24
06行业动态8 篇
- #7
- #105 ways AI Mode in Search helps you enjoy the real world
AI Mode in Search can help users host dinner parties by integrating with apps like Canva. Users can prompt Search to "Create a flyer in Canva for the party I’m hosting in two weeks. Make it modern minimalistic style," generating a custom design directly in search results. This allows for easy editing or downloading, enabling users to focus on other party preparations.
1 个来源 · 热度 32 - #115 ways to host the ultimate dinner party with Google Search
Google Search can help with dinner party planning, from creating menus to discovering dishes, allowing hosts to focus on guests. This tool aims to simplify the heavy lifting involved in organizing a dinner party, ensuring a more enjoyable experience for everyone involved. The official blog post from blog.google, published on July 28, 2026, highlights "5 ways to host the ultimate dinner party with Google Search."
1 个来源 · 热度 32追踪这条信号 - #13
- #14
- #19
- #23Apple overtakes Nvidia to become the world's most valuable publicly traded company, two years after it lost the crown, closing at a $4.9T market cap on Monday (Kalley Huang/New York Times)
Apple has surpassed Nvidia to become the world's most valuable publicly traded company, achieving a $4.9T market cap on Monday. This milestone comes two years after Apple previously lost the top position. The shift occurred amidst increasing concerns regarding the costs associated with the artificial intelligence boom, allowing Apple to slip past the chipmaker Nvidia.
0 个来源 · 热度 27 - #26