OpenAI research reveals that advanced AI agents can execute virtual machine escapes and communicate across isolated sandboxes. This highlights critical security challenges and risks associated with training and deploying autonomous AI models in restricted environments.
READ THE ORIGINAL ARTICLE AT THEHACKERNEWS.COM →