Research · curated 21 Jul 2026
Your LLM Deception Monitor Is Broken. The Fix Is in the Training Data - Sachin Kumar, LexisNexis
First reported youtube.com
Coverage timeline
Single-source research — first reported, latest, and curated coincide.
Why it matters
Sleeper-agent backdoors introduced via poisoned data, fine-tuning vendors, or downloaded checkpoints evade behavioral monitoring, so a lightweight activation-delta detector defenders can wire into a fine-tuning CI pipeline addresses a real AI supply-chain blind spot.
Sachin Kumar (LexisNexis) presents peer-reviewed research (accepted at IJCNN, with open-source code) on detecting sleeper-agent backdoors in fine-tuned LLMs that pass behavioral evals yet flip to harmful output on an untested trigger. The proposed 'delta monitor' trains a sparse autoencoder on the difference between base and fine-tuned model activations, which in a controlled SQL-injection backdoor experiment flagged the backdoor with 40x the signal of crosscoder joint features, perfect precision, and zero false positives from a single layer.