Skip to content

Scan record

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

Nov 1639 loaded

  1. Securing Multi-Agent Systems by Design

    PaperarXivThe item you came from

  2. An operational definition of opaque serial depth

    PaperarXiv

  3. Beyond Attack-Based Accuracy: Rethinking Adversarial Robustness Evaluation via Structural Gradient Alignment

    PaperarXiv

  4. Decoupling Forgetting and Utility: Lightweight LLM Unlearning via Orthogonal Null-Space Projection

    PaperarXiv: Jailbreaks, Prompt Injection, Poisoning & Defenses

  5. Has Abliteration Become Harder? Not Consistently: Direction-Based Ablation and ARA for Refusal Removal in Qwen3.8, Gemma 4, Olmo 3.1 and Mistral Small 3.2 LLMs and Their Predecessors in Heretic

    PaperarXiv: Jailbreaks, Prompt Injection, Poisoning & Defenses

  6. Constructing a steering vector via input optimization

    PaperarXiv: Jailbreaks, Prompt Injection, Poisoning & Defenses

  7. Alignment via Reflect-Reasoning: Learning How to Follow Principles

    PaperarXiv: Jailbreaks, Prompt Injection, Poisoning & Defenses

  8. FT Lex commentary: publishers are capable of a tough contest with big tech companies

    NewsFinancial Times · AIAI updates

  9. Former Nvidia chief Jensen Huang steps down in disgrace, dealing a blow to "open" models, with Anthropic and OpenAI benefiting

    NewsFinancial Times · AIAI updates

  10. Japan plans to regulate unauthorized commercial use of AI to replicate celebrities' voices and likenesses

    Newskmkj.krGovernance

  11. OpenAI Foundation board member says AI incidents will become the most likely cause of human death, warns of AI losing control

    NewsSegye IlboOpinion

  12. US Commerce Secretary Lutnick reportedly clears AI chips and advocates lax regulation, possibly involving family business interests in Southeast Asia

    Newsglobal.hk01.comGovernance

  13. Korean Association of Tax Accountants says practical training for trainee tax accountants will introduce assessments without AI assistance

    NewsFinancial News KoreaEvent

  14. Online post claims to have used AI to cheat into medical school; anxiety rises ahead of Korea's Suneung

    NewsSBSEvent

  15. Korean police seize devices including a micro camera in the AI cheating case involving the top scorer in the tax accountant exam, and are expected to summon him soon

    NewsSBSEvent

  16. Korean government to form interagency emergency team to prevent AI cheating on the College Scholastic Ability Test

    NewsKorea JoongAng DailyEvent

  17. Anonymous internet user posts claiming to have cheated with AI on Korea's Suneung as early as 2024, prompting calls to postpone the exam

    NewsThe Korea HeraldEvent

  18. Fed financial stability report warns of AI-related risks: forward leases and backstop guarantees for data center construction have reached about 3 trillion dollars

    News澎湃新闻 · 检索「AI风险」AI updates

  19. Jacob Helberg, U.S. Deputy Secretary of State for Economic Affairs, said the government needs to better explain the benefits of AI to the public

    NewsSemaforAI updates

  20. A LessWrong user attends a Boston AI protest with family, calling for regulation of AI

    NewsLessWrongGovernance

  21. Can LLMs Write Backdoors? An Empirical Security Study of Weight-Space Programming

    PaperarXiv

  22. Binarization as an attack surface of weightless neural networks

    PaperarXiv: 其他 AI 安全论文

  23. LLMs are vulnerable to malicious requests whose intent is hard to distinguish

    PaperarXiv: Jailbreaks, Prompt Injection, Poisoning & Defenses

  24. Representational geometry as a diagnostic tool for adversarial robustness

    PaperarXiv: 其他 AI 安全论文

  25. Forescout researchers used AI to discover security vulnerabilities in discontinued devices that no one is patching

    NewsHelp Net Security AIAttacks

  26. Western Cape Department of Infrastructure (DOI) and AIFOD explore responsible AI in infrastructure

    NewsWCGAI updates

  27. Amazon says its honeypot captured records of AI attack agents at work, with attackers using stolen API keys to drive attacks and search for more keys

    NewsTechnoEdgeEvent

  28. The Japanese government is considering establishing new penalties in the revision of the AI law for companies developing high-performance AI, to punish violations of orders

    NewsinnovaTopiaGovernance

  29. Top scorer in South Korea's tax accountant exam hands exam-related devices to police in AI cheating case

    NewsSeoul Economic DailyEvent

  30. Korean education YouTuber and lecturer call for conditional postponement of CSAT

    NewsChosunBizEvent

  31. Australia plans to establish a frontier AI safety regulator and mandate incident reporting

    NewsCSTIGovernance

  32. The top scorer in South Korea's certified tax accountant national examination has confessed in writing to police, detailing how he cheated using a miniature camera and AI

    NewsKorea JoongAng DailyEvent

  33. Kim Min-seok, leader of South Korea's Democratic Party: studying extending exam ban for AI cheaters from 5 years to 10 years

    NewsFinancial News KoreaGovernance

  34. Four days after Jensen Huang stepped down, talks of his return are reported, with over 150,000 people signing a petition supporting his return

    NewsSBSAI updates

  35. AI agent prompt injection and key exfiltration: Stroq open-source test, 36 attack scenarios, 0 penetrations

    NewsWeChat · Goumang Security LabTools

  36. Cursor affected by the Fireworks AI intrusion incident; a malicious version of the Claude Code extension was once listed on its marketplace

    NewsiThome (Taiwan)Event

  37. Uber: personal ID information of thousands of drivers leaked due to the Fireworks AI hack

    NewsiThome (Taiwan)Event

  38. The EU has always found a way out, but on AI it struggles to see a direction

    NewsPOLITICOGovernance

  39. Car owner says Claude Code used a $3 Arduino to crack and reverse-engineer the engine computer of a 32-year-old Jeep

    NewsCybersecuritynewsAI updates