[KaiserPro]

I think the chances of that data being released without a court order is fairly low. Ideally both the raw data, plus the product (actions based on the raw input, not the model/program) should be released, so that it can be included in everyone's dataset.

However, being that the sensors used are supposedly secret and innovative, along with the programs/models, I don't see the motivation for any AV company for releasing the data

[ptaipale]

Isn't NTSB usually releasing this kind of material after investigation? I do think it is reasonable that they do not release it before. The idea of NTSB looking at things is to analyse and find what to improve, not really to assign guilt (which appears to be the primary motivation just now).

[sokoloff]

(Assuming it's the same as aviation, with which I'm very familiar), NTSB factual reports are the only reports admissible into court. NTSB preliminary and final reports, which are the ones which contain findings or probable or contributing causes are statutorily barred from being admitted into evidence of other trials. (This is to encourage open participation from potential parties to future lawsuits.)

[FabHK]

Hugely sensible policy, I think, that contributes to the amazing safety record of aviation. I wonder whether it's the same for other modes of transportation.

[KaiserPro]

Good question. I'm not based in the US, so I don't know the ins and outs of federal or state level investigations.

In the UK I think the data is released to the coroner (assuming a fatality) but I'd have to research that to be sure.

[ptaipale]

I'm not in the US either, and my knowledge of the matter is largely from Mayday or Air Crash Investigations which runs on the National Geographic terrestrial TV channel here. But I would expect that this Uber case will get a somewhat similar treatment as airline crash investigations.