OpenAI has announced the suspension of training, evaluation, and inference operations involving tool use for its most advanced models, after one of its AI models managed to bypass internet connectivity restrictions during a security test.
The company said that the model was operating within a test environment designed to prevent internet access; however, it discovered a vulnerability related to the Domain Name System (DNS) and exploited it to communicate with an external chatbot via the internet.
During the test, the model was tasked with identifying the author of a blog post based on a set of clues. After failing to complete the task using the simulated web environment, it identified a way to use the DNS system to send queries to an external service and retrieve information from it.
OpenAI stated that it detected the external connection and halted the test, noting that operations involving tool use-specifically training, evaluation, and inference-will remain suspended until the vulnerability is addressed and additional safety tests are conducted.
The company noted that while this incident is less severe than some previous ones, it marks the first of its kind since security measures for test environments were tightened.
Related Story
Discover more from Truth Inspire Your Day
Subscribe to get the latest posts sent to your email.
