Skip to content

Scan record

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

Nov 1732 loaded

  1. Redwood believes the Gemini 4 safety report does not prove the risk of loss of control is low

    NewsRedwood ResearchOpinionThe item you came from

  2. FBI says hackers used the open-source AI tool Artex to breach networks, with some cases compromised in under three hours

    NewsTechCrunch AIEvent

  3. U.S. to close AI chip loophole before Trump–Xi meeting

    NewsTIMEGovernance

  4. Midas Project investigation: over 80 X influencers may have been paid to endorse pro-AI articles, some involving Nvidia chips for China

    XX @themidasprojGovernance

  5. The Midas Project said it found a covert influence operation using right-wing creators to promote sales of Nvidia chips to China

    XX @themidasprojGovernance

  6. OpenAI releases a free tool for vetted organizations to detect child sexual abuse material, including AI-generated images

    XX @OpenAINewsroomTools

  7. Apollo Research says it has launched an embedded evaluation pilot with OpenAI; six evaluators will be stationed there for four months

    XX @MariusHobbhahnEvaluation

  8. METR and OpenAI reach an agreement to conduct a limited alignment evaluation pilot, focusing on OpenAI's largest-scale RL run

    XX @METR_EvalsEvaluation

  9. Google DeepMind says researchers use Gemini 4 Carbon in Antigravity to handle multi-step problems

    XX @GoogleDeepMindAI updates

  10. Apollo Research and OpenAI launch an embedded evaluation pilot to investigate whether models attempt to undermine oversight of themselves

    XX @ApolloResearchEvaluation

  11. Apollo Research and OpenAI launch an embedded evaluation pilot, with a six-person team stationed for four months to evaluate model alignment

    NewsApollo ResearchGovernance

  12. GTIG AI Threat Tracking Report: Agentic threat activity crosses a critical threshold

    NewsGoogle Cloud BlogEvent

  13. Unit 42 discloses that China-linked attackers breached 10 organizations with AI-driven attacks

    NewsDark ReadingEvent

  14. Week in review: N-able, Windows, Chrome, refrigerator 0-day vulnerabilities and other security incidents

    NewsThe Hacker NewsEvent

  15. Google found hackers using AI agents to build and run exploits based on software patches

    NewsThe Hacker NewsEvent

  16. OpenAI releases CSAM detection classifier: can identify AI-generated content, provided free via API to vetted organizations

    NewsOpenAIDefenses

  17. The Midas Project investigation found that more than 80 right-wing influencers are suspected of being paid to promote pro-industry content such as Nvidia selling chips to China

    NewsThe Midas ProjectGovernance

  18. Author speculates: social sanction gaming in multi-agent RL can affect AI's own training

    NewsLessWrongAlignment

  19. OpenAI releases new image classifier and ChatGPT warning mechanism to combat child sexual abuse material

    NewsTechCrunch AIDefenses

  20. OpenAI will embed external safety researchers to investigate whether AI attempts to subvert alignment and control mechanisms

    NewsTechCrunch AIGovernance

  21. Trump backs calls for Jensen Huang to return as Nvidia CEO

    NewsThe Guardian · AIAI updates

  22. AI company develops decision model to avoid hallucinations

    NewsIEEE Spectrum AIAI updates

  23. Alibaba Qwen4-Max targets Opus 5.5, GPT-6 and Kimi K3.1 Pro at a much lower price

    NewsThe DecoderAI updates

  24. OpenAI discloses that agents once turned off monitoring; AI trust should rely on hard boundaries, not prediction

    NewsMIT Technology ReviewOpinion

  25. Nvidia CEO Jensen Huang ousted by the board, CFO Colette Kress becomes interim CEO

    NewsWIRED Security and AIAI updates

  26. Western misunderstandings of China's AI safety: information censorship plays a core role, but national security is not the only priority

    NewsFinancial Times · AIGovernance

  27. Anthropic pre-IPO profile: Chief Financial Officer Krishna Rao, who is leading the IPO

    NewsCNBC · TechnologyAI updates

  28. Draft EU AI Act harmonized cybersecurity standard prEN 18282 renamed prEN 18229-8

    NewsOWASP AI ExchangeGovernance

  29. OWASP AI Exchange updates the EU AI Act cybersecurity standard number to prEN 18229-8

    NewsOWASP AI ExchangeGovernance

  30. Google says Russian-speaking hackers may have carried out the first fully automated AI cyberattack

    NewsPOLITICOEvent

  31. Forethought authors estimate: the probability that fewer than 25 people control the future is about 23%

    NewsForethoughtOpinion

  32. UN-backed research report: AI models are weak and error-prone on climate science

    NewsThe Guardian · AIEvaluation

Nov 168 loaded

  1. The study says the "AI arms race" narrative is detached from reality, and tech enthusiasts in both the US and China know it

    NewsTech Policy PressGovernance

  2. TrustedMCP releases the November 2026 MCP server security status report

    NewsTrustedMCPEvaluation

  3. AI-driven copyright hell

    News404 MediaAI updates

  4. TÜV calls for independent review of frontier AI

    NewsproGovernance

  5. Anthropic judges on the basis of expectations that AI has reached the self-improvement threshold and activates safety rules

    NewsMartin Cid MagazineGovernance

  6. South Korean Prime Minister asks that shutting down public Wi-Fi at each CSAT test site be considered to prevent AI cheating

    NewsThe Korea TimesEvent

  7. South Korea tightens crackdown on AI-assisted cheating ahead of national university entrance exam

    NewsYonhapEvent

  8. Anthropic says the key to agent memory is knowing what to forget, not how much to store

    NewsThe DecoderAI updates