1sec.ai

Tag

#risk-assessment

Every item tagged risk-assessment, newest first.

3 items

researchOct 23

Strengthening our Frontier Safety Framework

Google DeepMind is strengthening its Frontier Safety Framework to better identify and mitigate severe risks from advanced AI models. The updated framework aims to improve safety and reliability as AI capabilities advance. Builders should expect more robust risk assessments and mitigation strategies from AI developers. This update reflects the growing importance of safety in AI development.

Key takeaways
  • Updated framework for identifying severe AI risks.
  • Aims to improve safety and reliability of advanced AI.
  • More robust risk assessments expected from AI developers.
DDeepMind#ai-safety#frontier-models#risk-assessment
otherApr 15

Our updated Preparedness Framework

OpenAI published an updated Preparedness Framework for assessing and mitigating risks from advanced AI capabilities. The framework aims to identify and address potential severe harms before they arise. It includes new risk categories and updated evaluation methods. You can use this framework to assess risks in your own AI projects.

Key takeaways
  • OpenAI updated its Preparedness Framework for frontier AI risks.
  • New risk categories and evaluation methods added.
  • Framework available for use in assessing AI project risks.
OOpenAI#ai-safety#risk-assessment#frontier-ai
researchJan 31

Building an early warning system for LLM-aided biological threat creation

OpenAI is developing a framework to assess the risk of LLMs aiding in biological threat creation. In a study with biology experts and students, GPT-4 showed only a mild increase in accuracy for biological threat creation. The results are inconclusive but provide a starting point for further research. You can use these findings to inform your own risk assessments and mitigation strategies.

Key takeaways
  • GPT-4 provides at most a mild uplift in biological threat creation accuracy.
  • OpenAI is developing a framework for evaluating LLM-related biological risks.
  • The study involved both expert and non-expert participants.
OOpenAI#llm-safety#biological-threats#risk-assessment