54 Matching Annotations
  1. Jun 2026
    1. We stand by this defense in depth strategy. It reduces the risks posed by Fable, making them comparable to the risks of existing models already deployed across the industry.

      大多数人认为发现新模型的漏洞意味着其风险高于现有模型,但作者认为通过深度防御策略,Fable的风险与现有模型相当。这挑战了人们对新技术风险更高的普遍认知,暗示新模型不一定比旧模型更危险。

    1. Anthropic singled out cybersecurity and biology as two domains where the safeguards may block responses, both areas widely considered sensitive topics for advanced AI systems.

      文章暗示了AI在特定领域的风险,但未详细解释为何这些领域被视为敏感。需要深入了解Anthropic的安全措施具体如何工作,以及这些限制是否足够全面,是否存在其他潜在风险领域。

  2. Apr 2026
    1. We separately evaluate GPT‑5.5 Pro in certain cases because we judge that the setting could materially impact the relevant risks or appropriate safeguards posture.

      大多数人认为如果两个模型使用相同的基础架构,它们的风险和安全需求应该相似,但OpenAI明确表示GPT-5.5 Pro需要单独评估,因为'设置可能显著影响相关风险或适当的安全措施立场'。这挑战了AI评估领域普遍认为的'相同基础模型的安全特性一致'的共识,暗示即使是微小的设置变化也可能导致显著不同的风险特征。

    1. Real-time monitoring of agent actions with a 12-category anomaly detection system derived from frontier model safety evaluations. Three-level alert system: PROHIBITED (immediate block), HIGH_RISK_DUAL_USE (human review), DUAL_USE (log and track).

      这种三级警报系统展示了AI安全监控的精细化程度,将代理行为分为不同风险级别,从完全禁止到仅记录跟踪。这种分类方法反映了AI安全中'双重用途'挑战的复杂性,即同一技术既可用于防御也可用于攻击。

  3. Apr 2025
  4. Aug 2024
    1. the conclusion must of course be, okay, so this is the risk  assessment we have. Let's, then we have to apply precaution. Precautionary  principle. Exactly. Uncertainty in science, which will always be there, should in my view,  always be. connected with a risk assessment.

      for - adjacency - precautionary principle - risk assessment - progress traps

      adjacency - between - precautionary principle - risk assessment - progress trap - adjacency relationship - Precautionary principle is really stating that we don't have enough knowledge and there can be a high risk - Even if there is low probability of occurrence, we must apply precautionary principle to avoid a progress trap

  5. Mar 2024
    1. Die Europäische Umweltagentur hat ihren ersten Klimarisiko-Bericht veröffentlicht. Von 36 Risiken erfordern 21 sofortiges Handeln, acht mit besonderer Dringlichkeit. Insgesamt sei Europa bei weitem nicht ausreichend auf die Risiken der globalen Erhitzung vorbereitet, die in Südeuropa am bedrohlichsten seien. Europa ist der von der Erhitzung am stärksten betroffene Kontinent. https://www.derstandard.de/story/3000000211032/eu-muss-sich-auf-katastrophale-folgen-des-klimawandels-vorbereiten

      Bericht: https://www.eea.europa.eu/publications/european-climate-risk-assessment

  6. Apr 2023
  7. Aug 2022
  8. Apr 2022
  9. Feb 2022
    1. Meaghan Kall. (2022, February 17). BA.2 risk assessment New this week is upgrading Immune Evasion—Amber 🟨 from low to moderate that BA.2 is antigentically different to BA.1 Unsurprising given the mutation profile, with BA.2 slightly more immune evasive than BA.1 on neuts studies https://t.co/n6DWtiRaNH [Tweet]. @kallmemeg. https://twitter.com/kallmemeg/status/1494100170195312646

  10. Jan 2022
  11. Dec 2021
  12. Oct 2021
  13. Sep 2021
  14. Jul 2021
    1. Gargano, J. W., Wallace, M., Hadler, S. C., Langley, G., Su, J. R., Oster, M. E., Broder, K. R., Gee, J., Weintraub, E., Shimabukuro, T., Scobie, H. M., Moulia, D., Markowitz, L. E., Wharton, M., McNally, V. V., Romero, J. R., Talbot, H. K., Lee, G. M., Daley, M. F., & Oliver, S. E. (2021). Use of mRNA COVID-19 Vaccine After Reports of Myocarditis Among Vaccine Recipients: Update from the Advisory Committee on Immunization Practices — United States, June 2021. MMWR. Morbidity and Mortality Weekly Report, 70(27), 977–982. https://doi.org/10.15585/mmwr.mm7027e2

  15. Jun 2021
  16. May 2021
  17. Mar 2021
  18. Feb 2021
  19. Oct 2020
  20. Sep 2020
  21. Aug 2020
  22. Jul 2020
  23. Jun 2020
  24. May 2020
  25. Apr 2020