LodeHQ
AI news for people who shipFree, by email

AI · alignment.anthropic.com · Saturday, 14 November 2026

Adding clean data can turn weak poison into a hidden backdoor

A few poisoned fine-tuning examples that do little alone can create a hidden backdoor when mixed with clean examples, so auditing poison in isolation may miss the risk. In the researchers’ setup, poison examples pair a task with a trigger, while clean examples remove that trigger; regularization then makes the model attribute the behavior to the trigger. They reproduce the effect in small Qwen3-8B, Gemma-3-4b-it, and MNIST experiments, not a production pipeline.

Read the source: alignment.anthropic.com ↗In the 14 November issue →