Skip to content

Scan record

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

Nov 1739 loaded

  1. OpenAI signs contracts with four independent evaluation organizations to conduct on-site safety evaluations

    NewsUnite.AIGovernanceThe item you came from

  2. The UK AI Safety Institute (AISI)'s pre-deployment evaluation model is under pressure; the author suggests expanding AI assurance functions

    NewsTech Policy PressGovernance

  3. Apple tightens macOS app permissions to limit data access in response to AI agent security risks

    NewsThe Hacker NewsDefenses

  4. WIRED 2026 Year in Review: Interview with New York City Mayor Zohran Mamdani on data centers, AI and democracy

    NewsWIRED Security and AIAI updates

  5. Out-of-control agents breach Hugging Face; why is its CEO Clément Delangue still bullish on AI?

    NewsWIRED Security and AIEvent

  6. Apple Watch's Live Rewind can record and rewind spoken conversations, and the experience is quite awkward

    NewsWIRED Security and AIAI updates

  7. Colombia's House of Representatives establishes a special AI committee to fill the AI legislative gap

    NewsCámara de representantesGovernance

  8. Superintelligence Task Force should include third-party vulnerability testing and disclosure in its assessment

    NewsCyberScoopGovernance

  9. AI guided two former Google employees to build the arms company Tiberius Aerospace, valued at more than $1 billion

    NewsWIRED Security and AIAI updates

  10. Japan and ASEAN leaders agree to cooperate on ensuring AI safety and support LLM development

    NewsNikkeiGovernance

  11. EU scientific advisory bodies propose four AI regulatory recommendations

    Newswedoany.comGovernance

  12. Cisco Talos discloses how hackers use ARTEX, PentAGI and HexStrike AI to automate cyberattacks

    NewsCybersecuritynewsEvent

  13. Microsoft's guide to defending against indirect prompt injection attacks: defense in depth with multiple layers of mitigations

    NewsMicrosoftDefenses

  14. NSFOCUS Technology's Fengyun Wei foundation model tops the CyberGym Model leaderboard with a 91.31% task success rate

    News绿盟科技技术博客Evaluation

  15. Amid fallout from Jensen Huang's resignation, Nvidia falls 0.56% premarket ahead of earnings

    NewsCNBC · TechnologyAI updates

  16. How VLMs plan their next step

    PaperarXiv

  17. Differentiable Exact Learning

    PaperarXiv

  18. Recovery Goes Too Far: Understanding and Fixing Risky Fallback in LLM Agents

    PaperarXiv

  19. Self-evolving dual-role deliberation for medication recommendation based on clinical records

    PaperarXiv

  20. ASCG: Adaptive Spatial Classifier Guidance for Precise Concept Suppression in Diffusion Models

    PaperarXiv: Jailbreaks, Prompt Injection, Poisoning & Defenses

  21. Polish social media misogynistic speech detection dataset: transferability of hate speech detection from generic to specific hate types

    PaperarXiv: Jailbreaks, Prompt Injection, Poisoning & Defenses

  22. Between the Lines: Contextual Investment Advice Detection in Multi-Turn Conversations

    PaperarXiv: Jailbreaks, Prompt Injection, Poisoning & Defenses

  23. South Korean prosecutors seek three-year sentence in second instance of Chinese national graduate student deepfake sexual crime case

    NewsYonhapEvent

  24. Guatemala launches multi-stakeholder discussions to jointly develop a national artificial intelligence strategy

    Newsprensa.gob.gtAI updates

  25. Irene Liu on why California leads US frontier AI regulation: from the vetoed SB 1047 to SB 53

    NewsThe EU AI Act NewsletterGovernance

  26. An overview of AI safety in California state politics

    NewsLessWrongGovernance

  27. Rumored November 17 global blackout: a mix of distorted facts and fabricated videos

    NewsRTVEEvent

  28. British actors will gain more digital replica rights under an AI protection agreement with producers

    NewsThe Guardian · AIGovernance

  29. Backdoor trigger words are clearly readable in fine-tuned weights

    PaperarXiv: Adversarial Attacks, Backdoors & Poisoning

  30. Conv-Reluplex: A Verification Framework for Convolutional Neural Networks

    PaperarXiv: Adversarial Attacks, Backdoors & Poisoning

  31. Forging a Bearer to take over a LiteLLM MCP session: reproduction and analysis of the OAuth2 passthrough authentication bypass

    NewsXianzhi Community (Aliyun)Event

  32. NATO AI cyber defense uses up to 300B OpenAI tokens per month

    NewsHelp Net Security AIEvent

  33. Google finds AI agents writing and launching exploits without human involvement

    NewsHelp Net Security AIEvent

  34. CuRSE-Bench: A benchmark for probing cultural blind spots in text-to-image safety evaluation

    PaperarXiv: Dangerous Capabilities & Safety Evals

  35. Between Rigor and Reality: How AI Safety Benchmark Developers Understand Benchmark Use and Maintenance

    PaperarXiv: Alignment, Deception & Oversight

  36. Manipulating Bayesian inference and decision-making through data poisoning

    PaperarXiv: 其他 AI 安全论文

  37. Robustness as a Transferable Skill: Cross-Task Adversarial Defense for Vision-Language-Action Policies

    PaperarXiv: 其他 AI 安全论文

  38. The Paradox of Provenance and Privacy: Can Private Cloud Compute (PCC) Prove That Apple Reference Images Are Not Used for Learning?

    PaperarXiv: 其他 AI 安全论文

  39. When tool names are not identity: measuring provider misbinding in MCP-based AI agents

    PaperarXiv: 其他 AI 安全论文