Skip to content

Scan record

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

Yesterday · Nov 1939 loaded

  1. NVIDIA NeMo Guardrails v0.25.0 released: adds agent tool call checks and a gateway

    NewsNVIDIA NeMo Guardrails ReleasesToolsThe item you came from

  2. The author reviews Christiano's "Prosaic AI alignment": says it was largely right, but misjudged supervised learning

    NewsLessWrongOpinion

  3. South Korea's Personal Information Protection Commission and Japan's PPC sign MOC to strengthen cross-border data transfer and AI privacy cooperation

    NewsPersonal Information Protection Commission (Korea)Governance

  4. When tech workers have to resign to sound the AI safety alarm, it's time to worry | Isabelle Ferreras

    NewsThe Guardian · AIGovernance

  5. Claude opens an AI factory: 38 models built in 7 weeks, racking up 55,000 downloads

    NewsWeChat · AI Era (Xinzhiyuan)AI updates

  6. X users say Nvidia's interim CEO announced early pay raises at an all-hands meeting

    XX @MinLiBuildsAI updates

  7. The Netherlands Bar Association published revised recommendations on AI use in the legal profession: non-binding, covering five main themes including procurement and information security

    NewsNederlandse orde van advocatenGovernance

  8. What might a new social contract between the United States and the tech industry look like?

    NewsFinancial Times · AIOpinion

  9. Anthropic's AI progress raises new safety concerns on the eve of its IPO

    NewsFinancial Times · AIGovernance

  10. Anthropic to IPO, becoming the latest trillion-dollar-market-cap company in U.S. stocks

    NewsFinancial Times · AIAI updates

  11. Wiz analysis finds malicious PyPI versions of Microsoft Agent Framework have caused 3 successful data exfiltrations

    NewsNeoTeoEvent

  12. Anthropic's stock began trading on Nasdaq at 111.50 dollars per share, with a market value of 2 trillion dollars

    NewsThe Guardian · AIAI updates

  13. California plans to strengthen AI safety regulation; government could order emergency shutdown of models and limit R&D compute

    NewsiThome (Taiwan)Governance

  14. FVScore: Specification-driven neuro-symbolic video evaluation

    PaperHugging Face Daily Papers

  15. Non-human identity security for AI agents: why intent-based access and behavior control is the missing layer

    NewsLasso SecurityOpinion

  16. Ooredoo Group concludes Digital Ecosystem Conference, forming draft Doha Principles to advance a trusted digital economy

    NewsOoredoo GroupGovernance

  17. The Australian Securities and Investments Commission (ASIC) makes AI harm to consumers a 2027 enforcement priority

    NewsMLexGovernance

  18. Google says misconfigured test server exposed autonomous AI attack operation; system contained Russian-language comments

    NewsEscudo DigitalEvent

  19. Google AI agent, after receiving a "never give up" instruction, attempted hacking and impersonating employees

    NewsThe ElecEvaluation

  20. Jensen Huang and Nvidia are not a honeytrap story but a national security story

    NewsFortuneGovernance

  21. On the eve of Anthropic's listing, Sam Altman reiterates OpenAI's listing plan, having previously paused it over safety considerations

    News澎湃新闻 · 检索「AI安全」AI updates

  22. Microsoft agent-framework PyPI package poisoned

    NewsStepSecurityEvent

  23. When figures and text disagree: evaluating autonomous visual evidence verification in scientific papers

    PaperarXiv

  24. Semantics-aware contrastive and explicitly disentangled representations for AI-generated video detection

    PaperarXiv

  25. Log analysis for LLM agents: a lifecycle-oriented survey

    PaperarXiv

  26. IIJ Global and VeriCerts launch advertising trust gateway PoC: verifying ad issuers and tampering via DID/VC

    NewsIIJ Global SolutionsAI updates

  27. LLM jailbreak detection based on contrastive prototype structure

    PaperarXiv: Jailbreaks, Prompt Injection, Poisoning & Defenses

  28. George Monbiot commentary: The UK growth plan forces through data centers nobody wants, betting that the AI bubble won't burst

    NewsThe Guardian · AIGovernance

  29. Researchers propose inducing monitorable reward hacking to compensate for the failure of gradient routing

    NewsLessWrongAlignment

  30. A LessWrong author posted about protesting Anthropic's IPO on November 19, calling for a ban on recursive self-improvement

    NewsLessWrongGovernance

  31. Shanghai AI Lab paper: switch to a different agent execution framework, and model score rankings may reverse

    NewsWeChat · Weijiqi LixinAI updates

  32. Tencent Zhuque Lab open-sources the pikit toolbox to systematically evaluate AI agent indirect prompt injection risk

    NewsWeChat · Weijiqi LixinTools

  33. Beihang and Alibaba teams propose the PASP framework, using multi-perception views and adversarial self-play to improve large models' ability to recognize disguised harmful content

    NewsWeChat · Weijiqi LixinDefenses

  34. Paper: under ANCHOR multi-turn attacks, refusal rates of 8 CLI agents drop to 0%

    NewsWeChat · Weijiqi LixinAttacks

  35. Xint discloses 5 Nginx RCE vulnerabilities, four discovered by automation

    NewsWeChat · SecureNexusLabEvent

  36. OpenAI is testing giving its autonomous agent Dots its own email address

    NewsiThome (Taiwan)AI updates

  37. Solomonoff induction based on reflective oracles

    NewsAI Alignment ForumAlignment

  38. CCN updates ENS information security policy and responsibilities and functions guidance, adding an appendix linking data protection and AI rules

    NewsCentro Criptológico NacionalGovernance

  39. Can you make a model believe false facts in the prompt?

    PaperarXiv: Alignment, Deception & Oversight