Tesla FSD Learns to Slow Down for Police [VIDEO]

Tesla’s FSD is trained on real-world driving and is increasingly reflecting informal road norms. Owners have recently observed that it now proactively slows down, changes lanes, and blends into traffic when it detects a police vehicle in a highway median.

Another instance of Tesla FSD detecting a police truck in the median (pretty far away) then slowing down and moving over to blend in with slower traffic. This was in Hurry mode, speed limit 70 mph. Love it! pic.twitter.com/uH6FLkI1xM

— Sam D. (@SamuelD2022) May 22, 2026

Blending In

According to a recent post on X, a Tesla driver described how the vehicle reacted after spotting a police vehicle in the highway median: FSD slowed and shifted over one lane to merge with the slower flow of nearby traffic. The post also notes the drive was in Hurry mode with a speed limit of 70 mph.

By matching the pace of the surrounding pack instead of overtaking near the officer, the system avoided drawing attention, mirroring what a cautious human driver would do.

The Power of End-to-End Learning

This human-like response is likely a product of Tesla’s end-to-end neural network and the way it improves over time.

Because FSD is trained on millions of miles of real-world data from actual drivers, it naturally absorbs some human driving tendencies. Since most drivers instinctively brake and move away from the fast lane when they see a vehicle stopped in the median, the neural network has learned that this is the expected, appropriate action.

Beyond Emergency Vehicles

While the neural network is explicitly trained to yield to active emergency vehicles with flashing lights, this median-avoidance behavior appears to apply to almost any stationary vehicle.

Whether it is a police cruiser running a speed trap, a broken-down commuter car, or a highway maintenance truck, FSD has learned that giving vehicles on the shoulder a wide berth and approaching more slowly is a safe and natural way to drive.