JFrog

495 malicious AI models were identified on Hugging Face.

18% of organizations have zero governance over their IDE or MCP servers inside developers' workflows.

97% of organizations claim they have certified model governance.

Secrets detection is active at just 28% of organizations.

45% of security and DevOps professionals say reviewing and hardening AI-generated code is now a major time drain.

The "Qix" campaign used 25 packages to compromise over 2.5 million downloads.

969 malicious AI agent skills were identified carrying high-impact payloads.

56 malicious extensions were identified on OpenVSX.

Malicious npm packages surged 451% year-over-year.

177,000 new malicious packages were detected across registries in the last year.

Over 48,000 new CVEs were disclosed in 2025, a 20% year-over-year increase.

Injection (CWE-74) occurrences grew 3,110%.

66% of analyzed CVEs had minimal real-world applicability.

Only 40% of organizations have adopted malicious package detection.

53% of organizations self-host models from sources where malicious payloads have been detected.