Detecting Strategic Deception Using Linear Probes — AI Alignment Forum