Martin Lewis: Suing Facebook left me shaking - it’s now admitted 1,000s of fake ads, here’s the latest

_________________________________________________________________________

Q2134  Chair: You have probably been aware since you have been in the UK of the issues raised by Martin Lewis about the use of his image and profile by fraudsters seeking to scam people into investing in schemes that they falsely claim he has endorsed. Have you seen anything on that?

Mike Schroepfer: I have seen news report on it, yes.

Q2135  Chair: He makes two points I want to ask you about, which I think are quite important. First, it is often beholden on him as an individual to complain to Facebook about the adverts but if they are not being served to him, certainly if they are being posted as dark ads, how can someone complain about an ad that they can’t see?

Mike Schroepfer: You have raised a couple of issues there and I think they are critical, so I am going to take each of them. The first is the core issue of fraudulent ads is a problem and it is not welcome on our platform, whether it is this case or others. We have both manual and automated systems to try to detect these systems, but these are people who are financially motivated to try to evade whatever systems we have in place. We use things to classify these ads in a certain way. They will abbreviate the name, so instead of Martin Lewis they will say M Lewis or they will misspell the name slightly such that even when we put in technical measures they will work around it. They keep running ads and they will get blocked and blocked until they figure out they can get one. This is a place where taking down the actor is often the most important thing.

In the case of Mr Lewis, he reported in the order of 50 ads to us. As a result of that we did a more extensive investigation using our technical tools, found thousands of other ads that were a problem, and took all of those down proactively. More importantly, we found the dozens of actors, people who are fraudulently advertising on the platform and took them off the platform. That obviously takes everything they have advertised and prevents them from advertising in the future. I think that is the important thing.

The second issue you raised is also an important one, again about transparency: how does he see these things? This is where in June you will basically be able to see every running ad on the platform. I want to highlight that I think we are trying to catch all of these things proactively. I do not want the onus to be put on people to go and find these things and people reporting. We are trying to get to a mode over time, particularly with technical systems, that can catch this stuff upfront. We want to get to a mode where people reporting bad content of any kind is the sort of defence of last resort and that the vast majority of this stuff is caught upfront by automated systems. I know that is true of authentic accounts, bad content on the platform; 90%-plus of nudity on the platform is flagged and removed by automated systems before any human ever sees it, which was not true five years because we did not have the technical capability. That is the future that I am personally spending my time trying to get us to so that people like Mr Lewis don’t have to spend time doing it because our systems do it for him.

Q2136  Chair: But Mr Lewis said he does not advertise himself. He does not use any adverts that have his image in. Why doesn’t Facebook use the facial recognition technology it already has to identify ads posted by fraudsters using his image to prevent them going up?

Mike Schroepfer: The first thing is that we have not had facial recognition enabled in the EU until just recently. It was not even a feature we could offer to our users, so we are just in the process of rolling that out, so we need to do that first to make sure we can offer that feature for consumers. We are investigating ways to do that. It is challenging to do technically at scale and it is one of the things I am hopeful for in the future that would catch more of these things automatically. Usually what we end up doing is a series of different features would figure out that these ads are bad, so it is not the picture, it is the wording. We will catch classes of ads and say we are pretty sure this is a financial ads and maybe financial ads we should take a little bit more scrutiny on upfront because there is the risk for fraud. This is why, for example, we took a hard look at the hype going around cryptocurrencies and decided that when we started looking at the ads being run there the vast majority of those were not good ads. We just banned the entire category and said, “You can’t run cryptocurrency ads because we think the likely harm to the consumer is high there”. When we can do things like that, combined with catching the actors—because this is ultimately a person who is running a fraudulent ad and get them off the platform—that is the really effective long-term defence of these things.

Q2137  Chair: You can run facial recognition at scale, clearly, as a business?

Mike Schroepfer: I am happy to dive into the details on this. I am not sure we have the systems yet to be able to do that, running this across 2 billion people for every single ad that is uploaded, to figure out—

Q2138  Chair: But you are running facial recognition across 2 billion users for their pictures and things. The idea is that you have facial recognition software and you could automatically tag on their pictures unless they have opted out of that.

Mike Schroepfer: The challenge here, and forgive me if I go into some detail, is that when any of these systems happen and you are trying to match, there is this ad running, is this particular—does it match, the larger the search space you use if you are looking across a large set of people, the more likely you will have a false positive, that two people tend to look the same and you will not be able to make automated decisions that said this is for sure this person. This is why I say that it may be one of the tools but usually what ends up happening is a portfolio of tools. Maybe it is something about the image, maybe the fact that it has Lewis in the name, maybe the fact that it is a financial ad, wording that is consistent with financial ads, and we tend to use a basket of features in this in order to detect these things. You are right, the thrust of this is that we are working hard to provide more technical means that move this from reactive to proactive, which means we catch this stuff at creation time and it does not run on the platform and harm a consumer, which is the most important thing, and it does not put work on anyone else to go and chase these things down. 

Chair: My concern over this section we have just discussed is that a lot of the tools do seem to work for the advertiser more than work for the consumer.

_________________________________________________________________________