trojan

Security researchers have identified hundreds of backdoored and malware-laced models in public AI registries. Most organisations pulling models from Hugging Face and similar platforms have no controls in place to detect them.