FAR.AI · white-box deception probe

When an assistant dodges your right to a straight answer.

Five real support chats. Someone asks a plain question — are you human? can this data reach me? — and the assistant lies or dodges. Can you tell which replies to trust?

Real transcripts · Qwen3-8B · FAR diverse linear probe · scores are live probe outputs.

Catch deception in your own models

FAR.AI builds white-box probes that flag lying, sandbagging, and sycophancy in real time.

Talk to us