Tesla Drops Safety Score To 80; Sends FSD Beta To New Testers

Maybe your like

Should Tesla Adopt NVIDIA’s Cooperative Steering? [VIDEO] January 21, 2026 By Karan Singh Share on Facebook

Not a Tesla App

One of the standout features of NVIDIA’s new Alpamayo autonomous driving platform, which is set to debut on the 2026 Mercedes-Benz CLA later this year, is something they call Cooperative Steering.

Unlike Tesla’s Autopilot or FSD, which typically function on a binary switch (the system is either on or off), NVIDIA’s system allows for a grey area. If a driver grabs the wheel to dodge a pothole or nudge the car over for a cyclist, the system does not disengage entirely.

Instead, it cooperates with the driver, accepting human input while maintaining lateral control, then seamlessly resuming when the driver lets go.

For Tesla owners used to the torque break disengagement, where fighting the wheel instantly kicks you back to manual control, this sounds like an interesting feature. While you can push the wheel slightly to nudge FSD or AP in the lane without full disengagement, the flexibility is much more limited than on NVIDIA’s platform.

Is this a feature Tesla should add, or is it for an era gone by? You can view NVIDIA’s cooperative steering feature in the video below:

The Case for Cooperative Steering

Users sometimes experience intervention anxiety. When the car hesitates or positions itself awkwardly, the driver has to decide. Do they let FSD figure itself out and risk curbing a wheel or worse, or do they take over and disengage Autopilot? A more cooperative approach helps to remove this friction. A driver could nudge the car over, and the vehicle can continue driving.

This could also be a helpful feature when FSD isn’t taking the route you’d like. Instead of disengaging, you could perform a turn, then simply let go of the steering wheel and let FSD continue.

In addition, autonomous safety research suggests that the most dangerous moment in semi-autonomous driving, like with FSD (Supervised), is during the handoff. When FSD disengages, the car instantly transitions from having torque on the wheel to zero resistance. If the driver isn’t ready for the weight transfer, this can lead to swerving. Cooperative steering helps to minimize this corrective action and maintain the safety net intact.

The Case Against

There’s a particularly simple case against Tesla incorporating this type of feature - the goal isn’t FSD (Supervised), it's FSD (Unsupervised). The current iteration of FSD is just a stepping stone towards true autonomy, and implementing new features will take time and engineering effort that could have been spent improving FSD.

Tesla AI’s philosophy is that any human input is wrong, and the end state is autonomy, so the steps along the way are less important. Implementing cooperative features could be seen as a step backwards - optimizing for a Level 2 driver-assist future, rather than a Level 5 robotaxi future. If the car becomes too reliant on human inputs, it may mask underlying failures in FSD.

There’s also a valid safety argument against cooperative steering: who is really driving? If the system provides resistance but still allows input, a driver might mistakenly believe the car will stop at a red light while steering around an obstacle. Tesla’s more binary system is designed to be crystal clear - if the blue tentacle and Self-Driving label are gone, you're in control. Blurring that line can lead to driver complacency.

For a Robotaxi, cooperative steering is irrelevant - there simply is no steering wheel. But for the consumer cars we drive today, NVIDIA’s approach offers a potentially better user experience in this regard.

However, we have to ask ourselves, how long will this still be an issue? FSD v14 showed us that Tesla is inching closer toward true autonomy, and along the way, we touch the steering wheel less and less.