Cyberattacks could push open AI platforms to restrict access, Hugging Face warns

Open AI platforms face a growing risk of cyberattacks because they combine open participation with valuable data and credentials, Hugging Face executive Irene Solaiman says. Recent incidents have prompted concern that platforms may restrict access, potentially reducing opportunities for independent researchers and hobbyists to contribute.

An October attack on OpenCode targeted users’ third-party API keys, while an AI agent swarm accessed cybersecurity evaluation answers on Hugging Face this summer. Fireworks also disclosed that attackers accessed some customers’ model weights, datasets and billing information, and temporarily disabled certain dataset functions. Hugging Face says it is increasing monitoring, rotating administrative access tokens more often and tightening network controls; GitHub has also strengthened some publishing and token security requirements.