Skip to content

Scan record

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

Nov 1538 loaded

  1. A long LessWrong post reflects on the concept of "circuits": the author says neural network computation needs a clearer ontology

    NewsLessWrongAlignmentThe item you came from

  2. Microsoft Research details a dormant-neuron backdoor method targeting agents

    NewsCASRAIAttacks

  3. OpenAI Foundation board member Paul Christiano warns: AI incidents could soon become the leading cause of human death

    NewsChosunBizOpinion

  4. Sankei Shimbun: Fireworks AI, a contractor of Notion used by a partner, suffered a cyberattack; Sankei iD was suspended and has now been restored

    Newssankei.jpEvent

  5. Data center construction stocks lose $4.2 trillion in market value as the AI trade boom unravels

    NewsFinancial Times · AIAI updates

  6. As former Nvidia head Jensen Huang imploded, human judgment fell short of its reputation

    NewsFinancial Times · AIOpinion

  7. Manufacturing intimacy and deferred mourning: emotional predicaments in the AI era

    News澎湃新闻 · 检索「AI风险」Opinion

  8. California lawsuit seeks to halt advanced AI development at OpenAI, Meta, Google, Anthropic, and SpaceX

    NewsEl ImparcialGovernance

  9. IISE researcher compares the two camps of the Hiroshima AI Process and WAICO and analyzes risks for Japanese companies

    NewsnoteGovernance

  10. AI Agent security scanning, MCP detection and workflow graphs: a hands-on breakdown of the open-source agentic-radar

    NewsWeChat · Goumang Security LabTools

  11. SemiAnalysis says large-scale smuggling of Nvidia AI GPUs into China may be the deeper reason for Jensen Huang's departure from Nvidia

    NewsTom's HardwareGovernance

  12. On the day Claude Fable 5.2 launched, Pliny leaked the Claude Code system prompt, 950,000 characters in total

    NewsBAAI CommunityEvent

  13. "Bring back Jensen Huang": Musk joins an online petition with more than 110,000 signatures

    NewsSBSAI updates

  14. Commenters say OpenAI's incident reports are the most systematic, but still do not meet the standard of full transparency

    XX @BethMayBarnesOpinion

  15. Beth Barnes criticizes Trajectory Labs' misalignment test remediation plan as excessive: double sign-off for each experiment creates too much friction

    XX @BethMayBarnesOpinion

  16. Nigeria's 2027 general election: preliminary voter register alleged to be full of AI-generated images, electoral commission announces dedicated department to monitor AI threats

    NewsDWEvent

  17. Altimeter and Atreides urge Nvidia board to keep open the possibility of Jensen Huang returning as CEO

    NewsThe Economic TimesAI updates

  18. FBI issues FLASH alert: open-source LLM agent penetration testing framework ARTEX weaponized by attackers for automated intrusions

    XX @AISecHubEvent

  19. Forescout used off-the-shelf AI models to discover vulnerabilities in critical infrastructure software

    NewsForescoutEvaluation

  20. Cloudflare proposes the Modern Security Model to address AI penetration testing orchestration frameworks

    NewsCloudflareEvent

  21. Podqi investigation: ChatGPT, Gemini, Perplexity recommend counterfeit retail websites

    NewsPodqiAttacks

  22. Jensen Huang resigns amid growing questions over extramarital affair; whether the security team knew remains unclear

    NewsTechCrunch AIGovernance

  23. Uber data breach: driver's license information stolen via AI company Fireworks AI

    NewsBaller AlertEvent

  24. U.S. Treasury Secretary Bessent accuses Commerce Secretary Lutnick of failing to add Nvidia partner Megaspeed to the Entity List

    NewsDeccan ChronicleGovernance

  25. vLLM 0.30.0/0.31.0 has a race condition vulnerability; upgrading to 0.32.0 fixes it

    NewsGitHub Security AdvisoriesEvent

  26. Teach and Grow: An agent-centric architecture for general robot learning

    NewsFailure-FirstAI updates

  27. OpenAI explains data disclosure limits under Japan's generative AI principles code

    NewsOpenAIGovernance

  28. AWS API MCP Server local deployment deny list flaw CVE-2026-125271, fixed in 1.5.1

    NewsWeChat · Oxo SecurityEvent

  29. Musk and other tech executives co-sign petition calling on Nvidia to reinstate Jensen Huang as CEO

    NewsTechCrunch AIAI updates

  30. Project Zero switches to a Full 90 vulnerability disclosure policy due to LLMs accelerating vulnerability discovery and exploitation, shrinking to 30 days in 2028

    NewsProject ZeroGovernance

  31. US Treasury Secretary Bessent accused Commerce Secretary Lutnick of failing to add Nvidia partner Megaspeed to the Entity List

    NewsDeccan ChronicleGovernance

  32. X user says Claude chat auto mode occasionally refuses operations, and after sending a message granting authorization it seems to be lifted

    XX @simonwDefenses

  33. Buck estimates the probability that AI companies will build superintelligence within a year at about one third

    XX @NeelNanda5Opinion

  34. Australian government publishes "What we heard" report: more than 1500 public submissions will shape AI standards

    Newspmc.gov.auGovernance

  35. Anthropic CEO Amodei acknowledged huge uncertainty about next year days before a record IPO

    NewsThe DecoderAI updates

  36. ESLint ends paid contributor program: AI disrupts open-source collaboration

    NewsSocketAI updates

  37. The Australian AI policy debate is in full swing, with Liberal Party senator McLachlan saying art must never be sacrificed on the altar of corporate profit

    NewsThe Guardian · AIGovernance

  38. Inside the attack loop: five AI attack frameworks that by default require no human approval

    NewsElasticEvent