There are two things. We have talked about one already, but let me just briefly underscore them. We are constantly changing and refining the way we identify fake accounts using machine learning. So again, in elections following the U.S. presidential election, we did identify tens of thousands of fake accounts for each of these electoral moments, and we put them down proactively. Again, independent studies confirmed that in the German election, this phenomenon of information being spread by fake accounts was not a driver in that election. We feel we are making progress on this. I would never want to make you think that we think we've solved it. This is a constant game of steady improvement and technical improvement. We need to do that .
More recently, in the lead-up to the mid-term congressional elections in the United States, which I alluded to earlier, we have made commitments to roll out a second phase of ad transparencies. Again, as I mentioned, Canada is the first country that we have tested “View Ads” in. We are going to bring that to the United States prior to the U.S. congressional elections. We're also going to have special additional measures for political advertising, whether it be for individuals or issue-based ads. For those ads, we have a few measures that we will implement to try to ensure the authenticity of the individual who is running these ad accounts. For example, we will make sure they upload a government I.D., and we will then confirm their address by sending them a piece of mail with a special code in it. We will launch those things first, in advance of the mid-term congressional elections. Our intention, as we've announced, is to roll this out globally afterwards.