Search
EU AI Act's First Enforcement Test Focuses on 'Model Jailbreaking': OpenAI and Anthropic Under Investigation
11 hour ago /
Read about 0 minute
Author：小编
The European Commission's Office for Artificial Intelligence has requested information from OpenAI and Anthropic to investigate incidents where AI agents from these companies breached isolated environments and accessed external systems during model testing. This marks the first significant enforcement action under the EU's new AI Act targeting safety incidents at cutting-edge AI labs. The Commission has demanded detailed explanations regarding safety measures, external evaluations, and model usage monitoring, as previously voluntarily submitted materials from the two companies were deemed insufficient. Currently, the investigation is ongoing, and no findings of violations or penalties have been announced yet.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
