Their internal privacy controls are extensive.
As a human, it is nearly impossible to access someone's data, even when debugging code/ML models/etc.
And when you do need to, it requires exhaustive oversight/approval, and _everything_ is audited.
Several years out, I no longer trust Google - not because I trust my former (immediate) coworkers less, but because the direction the entire advertising/data science industry is taking as a whole is deeply concerning to me. I disagree with almost all of it on principle, and am no longer comfortable supporting it by allowing unfettered collection of my personal data, regardless of who is doing the collecting or how much they promise not to sell my data.
(Companies were never "selling" data anyway - they were using it themselves, sharing it with their partners without an explicit sale, and otherwise doing things with it that I don't approve of which do not meet the strict definition of "sale".)
It makes me uncomfortable that the data is always there, and the direction of the business just needs to change. Perhaps they're not being profitable enough for wall street? And on a time scale of 10 more years, I'm sure there will be a number of "incidents" in which teams were given approval to use the data in unsavory ways.
> I'm sure there will be a number of "incidents" in which teams were given approval to use the data in unsavory ways.
This is my other concern, closely related to the first. Data companies (Google included) have a very different idea of what is "savory" w.r.t data usage. Not from a place of malice, necessarily, but innocence/privilege/not thinking about the consequences.
Let's say the engineers are building a data-using feature, such as one which takes Fitbit health data and links it to your medical record to recommend tests or interventions that might benefit you. Those engineers may only think about how many lives this will save - the benefits of sharing this data. Because there are some benefits, for some people, in that use case. The problem is when those engineers do not consider all the many ways that sharing could go wrong, and how many other people could be hurt. Discrimination, denial of insurance, stalking, etc.
Personally, I think it would be incredibly beneficial for most software engineers to spend time learning hacking and adversarial thinking. Teaching the people who build these features to think about how the features could, and will, be misused would likely help them build better, safer features. (/soapbox :) )
I lean to agree that it's more of a shift in public perception of Big Data companies. And hurting Google even when it might be one of the better players.
2) Those data are never going away, and may not (=will not) belong to Google forever,
3) Data that exists can be demanded by the government, justly or otherwise, and they will get it—private companies running a dragnet spy operation on everyone is only marginally better than the government doing the same directly,
4) Targeted ads are adversarial. So are ads in general, really—yes I know they can be useful but I'm talking in general, and in practice—but it's definitely more off-putting to be spied on then have that information used against you.
5) Google's incentives suck, their ethics suck, and I don't trust one bit that they won't do all kinds of nasty shit with the data they've collected the moment that looks like it's best for their bottom line.
To echo what someone else said further up the thread, the fact that they have the data is a liability to people. Policies and protocols can change over time. Just because they have great data access protocols in place now doesn't mean it will always be that way.
What happens when the temptation for more profit becomes too great? There are plenty of companies who have (unknowingly) killed their golden goose for short-term profit.
That may be true.
However, the quality of their internal controls has nothing to do with why I don't trust Google. I don't trust Google's official and company-sanctioned uses of my data.
The only problem is that if everything you say is actually true (it might only haven been true for you, or for a subset of folks working there), there is strictly no guarantee that it will last, especially now that all the people who did make Google into what is is culture-wise are long gone enjoying their [B|M]illions.
The next wave of leaders that are taking over have far less scruples and inherited a no-string-attached treasure chest (actually, a better analogy would be a amazingly well stocked armory) and will do with it whatever they damn please.
And what about "External privacy controls"?
I think that even fb devs can say such thing but...