Skip to content

Scan record

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

Today · Nov 2033 loaded

  1. Zenity says runtime analysis can only detect anomalous agent behavior, not stop it

    XX @ZenitySecOpinionThe item you came from

  2. New California AI safety report recommends emergency shutdown authority and a cap on compute for recursive self-improvement

    XX @j_asminewangGovernance

  3. The open-source project RVC enables real-time AI voice changing, with an official reported end-to-end latency of 170ms

    XX @MinLiBuildsAI updates

  4. Blogger @MinLiBuilds says Anthropic's IPO was priced at $83, with on-site bidding before the open at $105

    XX @MinLiBuildsAI updates

  5. Apollo Research proposes that third-party embedded evaluators review the monitoring infrastructure of frontier AI developers

    XX @ApolloResearchGovernance

  6. Anthropic CEO Dario Amodei delivers an address on the company's initial public offering

    XX @AnthropicAIAI updates

  7. AI Weekly #195: Zvi says he heard Anthropic is IPO-ing today

    NewsLessWrongAI updates

  8. The authors state that models on the same inference server can communicate via a side channel

    NewsLessWrongAlignment

  9. LessWrong author proposes six detection claims for embedded evaluation of AI monitoring

    NewsLessWrongOpinion

  10. Chrome creates Mirror Dimension: safely evaluating browser agents in a synthetic network

    NewsGoogle SecurityEvaluation

  11. Report says Meta agent Muse confuses conversations and leaks private data

    NewsTechCrunch AIEvent

  12. Anthropic completes record IPO

    NewsSemaforAI updates

  13. Anthropic lands on Nasdaq, opening at 111 dollars, 33% above the issue price

    NewsTechCrunch AIAI updates

  14. Gewirtz on AI risk as seen from Beijing

    News80,000 Hours PodcastGovernance

  15. Anthropic began trading on Wall Street

    NewsThe Verge AIAI updates

  16. TSMC and Samsung halt some processor shipments to China; AI cloud compute in third countries requires licenses

    NewsThe DecoderGovernance

  17. Palo Alto's earnings beat expectations and its stock surged, with a heavy bet on open-weight model security

    NewsCNBC · TechnologyAI updates

  18. Vals AI builds RollerCoasterTycoonBench: three of nine decision-making models achieve the operating goal

    NewsVals AIAI updates

  19. Goiás state police in Brazil disclose AI policing covers 200 cities and facial recognition expansion

    Newsgoias.gov.brAI updates

  20. Claude writes its own inference engine! Waiting for AI's first token, cut from 12 seconds to half a second

    NewsWeChat · AI Era (Xinzhiyuan)Evaluation

  21. Anthropic rings the Nasdaq bell, raising $81.34 billion

    NewsWeChat · AI Era (Xinzhiyuan)AI updates

  22. Google builds a giant "Mirror Dimension" internet to test Chrome AI agents

    NewsWIRED Security and AIDefenses

  23. Natalie Silvanovich: Project Zero vulnerability disclosure deadline to be phased down to 30 days

    Newstl;dr secAI updates

  24. Cloudflare report: after ARTEX was taken down, more than 50 framework families emerged, with daily requests rising 30-fold in six weeks

    NewsWeChat · LLM Security ResearchEvent

  25. Anthropic CEO Dario Amodei said on the eve of the IPO bell-ringing that AI risk is "very real" but avoidable (video)

    NewsThe Guardian · AIAI updates

  26. While ringing the bell at the Nasdaq listing, the Anthropic CEO reiterated that it will control the pace of AI development

    NewsTechCrunch AIGovernance

  27. Apollo Research proposes six embedded evaluation claims for AI monitoring

    NewsApollo ResearchOpinion

  28. In a search question-answering task, researchers demonstrate that reference-answer feedback poisoning can cause an agent to self-generate backdoor rules.

    NewsWeChat · Mo'anjuAttacks

  29. Polish Senate to hold press conference launching national citizens' assembly on a safe internet

    Newssenat.gov.plGovernance

  30. Microsoft: Security in the AI era rests on the fundamentals — clear technical debt, treat agents as insider risk, defend at machine speed

    NewsMicrosoft On the IssuesOpinion

  31. ASIC announces 2027 enforcement priorities, will target consumer harm related to cost of living and technologies such as AI

    NewsASICGovernance

  32. OWASP AI Exchange updates agent overview based on SecureW interview: security scope should be defined by system properties rather than labels

    NewsOWASP AI ExchangeTools

  33. Large-scale system failures occur in succession in the United States; experts say AI-accelerated vulnerability discovery forcing emergency fixes may be the main cause

    News澎湃新闻 · 检索「AI风险」Event

Yesterday · Nov 196 loaded

  1. The author writes an article criticizing Anthropic for not living up to its own ideals

    NewsLessWrongOpinion

  2. AI and US elections: three takeaways from the NYU CSMAP and Knight-Georgetown workshop

    NewsTech Policy PressGovernance

  3. List of meetings and events hosted by China as host of APEC 2026

    News澎湃新闻 · 检索「人工智能安全」AI updates

  4. Biological data matters to the Genesis Mission: CSET report identifies five categories of data challenges

    NewsCSETAI updates

  5. Kuluttajaliitto warns: AI makes fake online stores harder to identify; check six tips before ordering

    NewsKuluttajaliittoEvent

  6. Meet Shepherd: a meta-agent that keeps runaway AI agents in check

    NewsIEEE Spectrum AIDefenses