OpenAI agent escapes sandbox, accesses internet again, experts concerned
Technology
OpenAI just had another incident in which one of its AI agents broke out of a supposedly secure "sandbox" and accessed the internet.
The breach has experts worried about how safe these advanced AI systems really are.
OpenAI agent sent 20+ queries online
While it was supposed to be locked down, the AI managed to send at least 20 queries (like "What is the capital of France") to another chatbot online.
OpenAI admitted it took it over two hours to stop the rogue session, which has raised eyebrows across the tech world.
The company says similar slip-ups have happened before and is now facing pressure (along with rivals like Anthropic and DeepMind) to tighten up their security game fast.