You raise several good points, but a few clarifications here:
- In our software that we deploy to police departments, generally the origin of any piece of information is tracked, as well as when it was entered and by who. The intention with this is to prevent this (and other) kinds of abuse. Malicious users can still abuse the system and do a parallel construction e.g. outside of our system or maybe in a way that the two actions (finding non-admissible data and finding admissible data) seem unrelated, but at that point they would be spending quite a bit of time and criminal energy on this. I don't think this would happen commonly for a variety of practical reasons (but ultimately there's no way to completely prevent it).
Ultimately trust in the government and that law enforcement (in the various shapes and forms it comes in) is a force for good and prevents many bad things from happening every single day, is at the heart of the palantir philosophy. If you fundamentally think the government is evil, palantir would probably not be the workplace for you. But once you see all of the bad things LE can prevent thanks to our software, it becomes pretty easy to believe in, even if individual bad actors exist (and always will), like racist cops, cops that abuse their power, etc.
- If the organization is acquiring warrantless surveillance, then this generally means that there is the legislative base for them to acquire the data in the first place. What this looks like varies from country to country, some countries are much stricter with the regulations than others. So for instance in some countries it is not allowed to get information from a suspects public facebook profile, whereas in other countries this is considered A-OK. Especially in europe these things have recently become much more strict with GDPR (and similar rules that apply to government entities). You'd think that be a downside to palantir (less data = less value?) but actually it's a big business opportunity, because our software is the only thing available that is even remotely close in this space to having enough access control 'finesse' that it can enable organizations to be compliant to this law. So this is an area (and competitive advantage) we are investing a lot into -- and typically organizations go from 'completely non-compliant' when we arrive there (e.g. never even deleting data like license plates that are supposed to be deleted after 3 months etc, audit-logging any searches investigators do, limiting search scope, ...) to being fairly or even fully compliant.
Now my personal opinion is that everyone should be able to enjoy great privacy and control over their PII just like citizens of the EU do, so if you live e.g. in the US other another country where the law is lax, you should consider taking political action to change the situation.
Of course we never endorse or support any workflows that are in any way unconstitutional, and we have terminated relationships before with very big government agencies when we had doubt about whether our tools would be used for unlawful purposes.
- The "big data" thing is a bit of a misunderstanding I think -- we have "big data" tools too, but usually these are not of interest to local law enforcement. If you check out our youtube channel you will find some (atrociously old!) videos of the tool that's popular in LLE (https://www.youtube.com/watch?v=yMv3TBxulu4 for instance). This tool is in fact often used with just hand-entered information -- analysts create objects (e.g. persons, links between them) in their investigation. So the data really isn't that big and mostly hand-curated, and the tool works fine without mass data collection (better than with, perhaps). Most of what local law enforcement does with 'big data' tools is to generate reports like crime statistics.
"Police departments don't just have random chunks of data lying around containing large numbers of connections between large numbers of people" -- actually, they really do! Any police department of any size that has existed for a while will have a database with millions of convictions, suspects, court-cases etc in them. Usually on some crufty mainframe or in some crufty old SQL database that contains a lot of terribly inconsistent data (dead links, data duplication, etc.) Also sometimes some of this data is shared between countries, states etc. You can find some videos on youtube of our CEO talking about the challenges of data integration and such.
Ultimately of course there is a general statistical trend (belief?) that 'more data is better', because if there is more and more complete data, you have a bigger chance of finding that connection between e.g. a terrorist and some billionaire who might be funding said terrorists, etc. At least up until a point (at which the data probably becomes just too noisy/hard to deal with, because every individual piece of information carries very little meaning. At least the NSA seems to subscribe to this belief according to their public statements, and I'd think if anyone knows about this stuff, it's them) and assuming you have the necessary CPU power and talent (data scientists) to actually do something useful with this data. Local law enforcement orgs like police generally lack the latter. In police departments, most users are only sophisticated enough to run searches for things like names, SSN, number plates etc, really.
- We generally know a lot about how our software is used at most of our installations, and almost always actively take steps to prevent abuse (auditing etc). Most government organizations will not let us see things like what their analysts exactly searched for (because that is obviously sensitive information) but we do a lot of work to help organizations prevent abuse and insider threats at a higher level. As mentioned above, we have taken action in the past in situations where we suspected abuse. It's on a case-by-case basis, of course -- imagine if there is e.g. one analyst abusing the system, then that analyst getting fired/reprimanded would probably be sufficient action (obviously this is not in our responsibility), but if we suspect systematic abuse or we see signs of repeated abuse without repercussions, then we might pull out completely.