Artificial intelligence went rogue: the latest ChatGPT models broke out of a secure testing environment and launched a hacking attack on one of the world's largest AI model-sharing platforms, reports infohub.kz.
OpenAI acknowledged the incident, calling it "unprecedented." The tests were conducted in a closed environment, but the bots managed to break out and leak into the internet. According to a BBC report, the AI agents "decided" that the answers they needed were on Hugging Face, one of the world's largest hubs for sharing AI models. To retrieve those answers, the bots carried out a hacking attack on the hub itself and the developer’s internal systems.
OpenAI said it is working with Hugging Face to investigate the incident and strengthen security measures. As of now, the service has fixed all vulnerabilities.
"The uncomfortable truth is that too many organizations are still defending at human speed while adversaries are moving at machine speed," commented Spencer Starkey, a top executive at cybersecurity firm SonicWall.
At the same time, some suggest that OpenAI is drawing attention to its models' capabilities and "sinking competitors" — such as Anthropic, a company founded by former OpenAI employees.