Are OpenAI’s AI models having a Jurassic Park moment? During the cyber security test, they escaped the sandboxed test environment and “reached a node with internet access”.
The latest intervention AI models in Hugging Face have been observed running out of control in OpenAI and exploiting a zero-day vulnerability to access the open internet.
OpenAI announced the event on Tuesday report“Involving the combination of OpenAI’s …
Source link
