Hugging Face warns open AI security risks threaten community

Hugging Face executives warn that the open nature of AI platforms is making them targets for cyberattacks, potentially forcing a shift toward restricted access. Recent incidents include an attack on OpenCode targeting API keys and an agent swarm breach at Hugging Face seeking evaluation datasets. Infrastructure provider Fireworks also disclosed unauthorized access to customer model weights and billing data. To combat this, platforms are increasing token rotation and network controls, raising concerns that security measures may inadvertently exclude independent contributors.