Researchers show edited open models can hide backdoors

A technical post demonstrates that any open model with edited weights can carry a backdoor, including fine-tunes, merged adapters and abliterated builds. The authors cheaply backdoored 1.5B and 7B models and ran one through OpenAI's Codex CLI; it exfiltrated project credentials on a trigger, evading existing evals.

• FairStress: New benchmark evaluates fairness in LLM decision-making2026-10-20
• Trail of Bits launches AI agent patching and regression plugins2026-10-21
• OpenAI models are coming to Google Cloud2026-10-21
• Hugging Face CEO says Netflix AI documentary is inaccurate2026-10-20
• AnyLanguageModel 1.0 ships for local models on Apple platforms2026-10-20