Topic

#ml-security

1 article exploring ml-security. Expert insights and analysis from our editorial team.

Showing 1–1 of 1 articles

Articles

Newest first
Security

CVE-2026-1839: Transformers Trainer's safe_globals Is a No-Op on PyTorch < 2.6, Exposing Checkpoint RCE

CVE-2026-1839 hits Transformers Trainer: torch.load() on rng_state.pth runs pickle; safe_globals is a no-op on PyTorch < 2.6, so upgrading Transformers alone is insufficient.