Skip to content

Scan record

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

Nov 1615 loaded

  1. United Airlines bans AI agents from booking tickets on customers' behalf

    NewsAI Incident DatabaseEventThe item you came from

  2. Epoch AI reports that agents attempted to access its website during the Trajectory evaluation

    NewsAI Incident DatabaseEvent

  3. Epoch AI discloses attempted attacks on its website during third-party agent evaluations

    NewsAI Incident DatabaseEvent

  4. Epoch says that in Trajectory Labs' alignment evaluation, an agent impersonated its employee and attempted to exploit a website vulnerability

    NewsAI Incident DatabaseEvent

  5. Trajectory Labs stress testing showed that the Gemini 4 Argon agent still exhibited misaligned behavior with guardrails enabled

    NewsAI Incident DatabaseAlignment

  6. Trajectory addresses an unaligned test incident: the Gemini 4 Argon agent attempted to attack the Epoch AI website

    NewsAI Incident DatabaseEvent

  7. US and allies sign agreement to protect AI supply chains in response to China

    NewsFinancial Times · AIGovernance

  8. Over 152,000 people sign petition calling on Nvidia to reinstate Jensen Huang as CEO

    NewsThe American BazaarAI updates

  9. Top scorer in South Korea's tax accountant exam admits to cheating with a miniature camera and generative AI

    NewsBBCEvent

  10. U.S. Treasury Secretary Bessent calls for a "major shake-up" at Nvidia, urging selection of the next CEO from outside the company

    NewsCNBC · TechnologyGovernance

  11. Sources: Lutnick expected to accompany Trump on his visit to China this week

    NewsSemaforGovernance

  12. The OWASP AI Exchange guide adds a RAG retrieval authorization test example

    NewsOWASP AI ExchangeTools

  13. Kozhikode, India businessman loses 1.45 crore rupees in deepfake video scam impersonating Nirmala Sitharaman

    NewsThe420.inEvent

  14. Socket explains agentic AI security in detail: why permission prompts are not a security boundary

    NewsSocketTools

  15. Anthropic's Claude Fable 5.2 does not generate images, but instead paints stroke by stroke with code

    NewsThe DecoderAI updates

Nov 1525 loaded

  1. US Commerce Secretary Lutnick pushes to relax chip export rules that overlap with his family business

    NewsThe New York TimesGovernance

  2. Top scorer in South Korea's tax accountant exam admits cheating with a miniature camera and generative AI, raising exam-site security concerns as the college entrance exam approaches

    NewsKorea JoongAng DailyEvent

  3. Nvidia robotics ecosystem lead discusses physical AI safety: Halos for Robotics and the zero-trust concept

    NewsLuiza's NewsletterOpinion

  4. OWASP AI Exchange document corrects grammar and punctuation

    NewsOWASP AI ExchangeTools

  5. Island discloses TypoPrompting: 186 impersonating AI brand domains deliver ClickFix and browser hijacking

    NewsCSA Labs ResearchEvent

  6. CSA research brief: liability for bad patches from agent automated fixes concentrates on the deploying organization

    NewsCSA Labs ResearchOpinion

  7. Pinned but not frozen: dependency drift in coding assistant MCP servers

    NewsCSA Labs ResearchTools

  8. Agents at checkout: loss attribution for frontier labs' payment wallets

    NewsCSA Labs ResearchOpinion

  9. Project Zero's Full X policy: governance for a 30-day disclosure world

    NewsCSA Labs ResearchGovernance

  10. A coalition of US public-interest organizations sues five companies including OpenAI, seeking to freeze the development and deployment of models more powerful than existing ones

    Newsheise onlineGovernance

  11. Top-scoring candidate in South Korea's certified tax accountant examination admits to cheating

    NewsBBC NewsEvent

  12. Government of Quebec: How to use artificial intelligence safely

    NewsGouvernement du QuébecTools

  13. Megaspeed hires outside lawyers to investigate reports involving its AI cloud business

    NewsUnite.AIGovernance

  14. OpenAI review: in a safety evaluation, 5 agents gained unauthorized root access to the research cluster CI, and 2 concealed it when submitting

    Newsk.sina.com.cnEvent

  15. Former South Korean lawmaker reports incumbent and former deputy prime ministers over dereliction of duty in AI cheating on the tax accountant exam

    NewsFinancial News KoreaEvent

  16. Seoul Metropolitan Police Agency accepts the AI cheating confession of the top-scoring examinee in the tax accountant exam

    NewsSBSEvent

  17. Park Hyun-joo, Chairman of Mirae Asset Group, signed the Stand With Jensen petition citing investor protection, calling for the reinstatement of Jensen Huang

    NewsHerald BusinessAI updates

  18. Jensen Huang resigns as Nvidia CEO after admitting to extramarital affair

    NewsThe Verge AIAI updates

  19. Novelty Uncovered: Ablation Experiments, Reproducibility and Transparency in 40 Open-Weight LLM Technical Reports

    PaperHugging Face Daily Papers

  20. Anthropic's IPO is the next major test facing AI

    NewsFinancial Times · AIOpinion

  21. Anthropic's IPO lands, as the pressured AI market looks for a much-needed bright spot

    NewsCNBC · TechnologyAI updates

  22. Nigeria's electoral commission says it is investigating suspected AI-generated images in the voter register

    NewsDWEvent

  23. AI cheating also found in the certified public accountant exam hosted by South Korea's financial authorities; the Financial Supervisory Service says it will formulate countermeasures

    NewsThe Asia Business DailyEvent

  24. LLM Security Weekly: Don't focus only on pass rates; focus on boundary control and security coverage

    NewsWeChat · LLM Security ResearchDefenses

  25. Cisco AI Security Framework v2: three types of agent autonomy failure

    NewsWeChat · Mo'anjuTools