A new AI model was used to hack an AI platform.
China's new open-source model, Kimi K3, launched on Monday.
Within a day, people reportedly used it to bypass the usage limits on Higgsfield, making the AI video platform free.
Instead of hiding the issue, Higgsfield publicly confirmed the exploit and said it would keep the platform free while fixing the problem.
It's another reminder that powerful AI models can be used for both innovation and attacks.