If it is operating mechanically, then it is following a process chosen by the developers who wrote the code. They work for the company, so the consequences are still the company's responsibility.
The car is following a process chosen by Mercedes' engineers (to go forward when the user presses the accelerator.) The newsfeed is likewise following a mechanistic process driven by user input (they wouldn't be showing misinformation if users weren't uploading and sharing them.)
If the Mercedes infotainment screen had shown you a curated recommendation that you run them over, prior to you doing so, they very possibly would (and should).
Would Mercedes be liable if the car (i.e. the algorithm) decided to run over someone?