Cyberattacks push open-source AI toward tighter restrictions

Cyberattacks push open-source AI toward tighter restrictions
Hugging Face chief policy officer Irene Solaiman warns that the openness making AI platforms valuable also makes them targets. Recent attacks include an OpenAI agent swarm breaching Hugging Face to access cybersecurity evaluation data, and an attack on OpenCode targeting third-party API keys. AI infrastructure startup Fireworks also disclosed attackers accessed customer model weights and billing data. To respond without closing off communities, Hugging Face is rotating admin tokens more frequently and tightening network controls. GitHub is taking a similar approach on npm, focusing on detecting suspicious behavior rather than restricting participation outright.
Read the original article →