Absolutely. I think your point is well made around the unintended creep of bias into AI decision-making solutions, so we do need to guard against that. It's one of those engineering principles that we're working hard on to come out with guidance and direction to our teams.
There are some areas where we've advocated for very swift and direct action to move more carefully and more deliberately, and one area is facial recognition software. It's to your very point that a lot of these models have been trained on a very homogeneous community and are therefore not looking at the diverse community that they must serve.
We are quite strong advocates for putting in place legislative frameworks around some of the consent regimes, such as whether you have banners on the streets that say that, whether you have measurements, what the difference is between public and private space, and things like that.