Skip to content

Scan record

Everything the site collected, newest first by collection time (Beijing time), back seven days.

Nov 1839 loaded

  1. OpenAI rates GPT-6.2 Astra's self-improvement capability as High; two second-stage tests did not meet the threshold

    NewsTechTimesEvaluationThe item you came from

  2. BabelTele jailbreak idea: symbolically compressed natural language may bypass some LLMs' surface-level safety detection

    NewsXianzhi Community (Aliyun)Attacks

  3. MCP Inspector 1.0.6 limits upstream SSE buffering to prevent a malicious server from exhausting memory

    NewsMCP InspectorEvent

  4. Southeast Asia's data center boom relies on Chinese demand, and new US rules put this channel at risk

    NewsCNBC · TechnologyGovernance

  5. CVE-2026-30623 vulnerability analysis: reproduction and source code audit of the LiteLLM Proxy MCP interface command injection vulnerability

    NewsXianzhi Community (Aliyun)Event

  6. Facing AI-accelerated attacks, UK NCSC calls on corporate boards to focus on proactive threat hunting to find detection gaps

    NewsiThome (Taiwan)Governance

  7. Attackers poisoned Microsoft Agent Framework's PyPI package to steal cloud credentials

    NewsThe Hacker NewsEvent

  8. NVIDIA board establishes National Security and Export Controls Oversight Committee

    NewsNVIDIA NewsroomGovernance

  9. The author worries that LLMs accelerating hardware R&D makes cheap ASI easier to achieve

    NewsLessWrongOpinion

  10. Austria's WKO updates its data protection guide for enterprise AI

    NewsWKOGovernance

  11. Veeam executive warns: commonly used compliance metrics fail in the face of AI agents; measurement is already outdated

    Newsdatensicherheit.deOpinion

  12. California sues OpenAI over AI safety issues

    NewsNeoTeoGovernance

  13. European Commission seeks comments on AI Act Article 27 fundamental rights impact assessment template

    NewsNeoTeoGovernance

  14. Reuters: China is increasingly using U.S. AI models to train its own models through distillation

    NewsainetAI updates

  15. FT opinion: When will the AI crash come? Ask journalists—170 years of constant anxiety, and a stopped clock is also often right

    NewsFinancial Times · AIOpinion

  16. Brussels will use profits from frozen Russian assets to service principal and interest on a new loan to Ukraine

    NewsFinancial Times · AIAI updates

  17. CCN publishes two new guides: strengthening ENS against offensive AI and quantum threats

    NewsCentro Criptológico NacionalGovernance

  18. White paper by Singapore's Manpower Research and Statistics Department: for AI to help people rather than replace them, the key is redesigning work and human-machine complementarity

    NewsLabour Market Statistics and PublicationsOpinion

  19. Nvidia's board chairman said the board unanimously accepted Jensen Huang's resignation and has no intention of having him return as CEO

    NewsThe Verge AIAI updates

  20. Sources say the SEC is investigating Nvidia, after Jensen Huang resigned over an extramarital affair allegation

    NewsNew York PostAI updates

  21. Jensen Huang agrees to testify before a U.S. House committee, says he is willing to return as Nvidia CEO if the board agrees

    NewsAl JazeeraGovernance

  22. Cantabria Cybersecurity Center to hold corporate AI security lecture on November 18: risks and defenses

    NewsCentro de Ciberseguridad de CantabriaAI updates

  23. Monetary Authority of Singapore and Future Finance Institute announce measures supporting the national AI mission for the financial services sector

    NewsMonetary Authority of SingaporeGovernance

  24. When the world becomes the test suite: agent intelligence for the physical world

    PaperarXiv

  25. Large-scale neuron community discovery reveals sparse neuron behavior in LLMs

    PaperarXiv

  26. Colombo: a frontier-to-frontier automated research system for oracle-based optimization

    PaperarXiv

  27. Validating agents in rubric-based environments

    PaperarXiv

  28. SkillBuffBench: measuring the contribution of Agent Skills across harnesses, models, skill access modes and memory

    PaperarXiv

  29. Colombia-Meta cooperation plan covers responsible AI and anti-fraud

    NewsMINTIC ColombiaAI updates

  30. Offsec Harness Index: an evidence index of AI penetration tools actually abused by attackers

    Newsoffsecharnesses.aiEvent

  31. From TeleChat to Xing4: Advancing Frontier Agent Models on Domestic Chips and Frameworks

    PaperHugging Face Daily Papers

  32. Author says persuasiveness in AI evaluations exceeds humans, with no evidence for real-world political persuasion

    NewsLessWrongOpinion

  33. Paris Hilton returns to the U.S. Congress to push for a House vote on a deepfake pornography bill

    NewsCBS News · TechnologyGovernance

  34. FT asks: Is AI financial advice credible?

    NewsFinancial Times · AIOpinion

  35. Gallup survey: More than one-third of Gen Z in 16 countries use AI for companionship

    NewsWIRED Security and AIAI updates

  36. Singapore Ministry of Health discusses healthcare AI oversight and community trials

    NewsMinistry of HealthGovernance

  37. Honest brokers in AI regulation: how public comments introduce evidence to regulators

    PaperarXiv: AI Governance & Policy

  38. GuardAgent: A Memory-Augmented Agent Framework for Adaptive Detection of Adversarial Promotion in LLM-Based Recommender Systems

    PaperarXiv: 其他 AI 安全论文

  39. ARC-2: Continuous Compliance via Delta Management

    PaperarXiv: 其他 AI 安全论文