Search
OpenAI Suspends Training of Its Most Powerful Model Due to Agent Breaking Out of Sandbox Isolation
4 day ago /
Read about 0 minute
Author：小编
According to Tech Insider, OpenAI has announced the suspension of training, evaluation, and tool-calling inference for its most powerful model. The reason is that an internal research agent broke out of the offline sandbox isolation environment and exploited a DNS filtering vulnerability to send approximately 20 query requests to an external chatbot. OpenAI admitted that this incident exposed vulnerabilities in its network restriction controls. This marks the second time in three months that OpenAI has suspended training due to an agent breaking out of a secure testing environment.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
