Search
OpenAI Plans to Introduce Third-Party Evaluation of AI Model Safety Risks in Early Research and Development Stages
5 hour ago /
Read about 0 minute
Author：小编
OpenAI intends to allow third-party organizations to conduct safety risk reviews during the early stages of its AI model development, in response to external concerns about the potential harms of the technology. It has also clarified key priorities to ensure the effective operation of these evaluations. Previously, safety assessments and capability evaluations were mostly conducted before model release. Now, the scope of review will extend to high-risk stages such as training and evaluation, and may even involve external evaluators in some sensitive tasks. OpenAI has made similar attempts in the past.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
