Mr. Ronan Costello:

At the Twitter transparency centre, available at ,about every six months or so we update that site with the latest figures on our enforcement. That would include all violations of Twitter rules and the information is broken down by rules. If we were talking about platform manipulation and spam, for example, which would be co-ordinated inauthentic activity, bots if one likes, those kinds of figures could be found there every six months, depending on the nature of the report.

This figure is not up to date but as an example, around 18 or 24 months ago, we were challenging between 8 million and 10 million accounts globally each week. Challenging means that something about the sign-up process with particular accounts was setting off flags. I mentioned earlier that we are leveraging machine learning to be better at detecting behaviour-based signals. Machine learning flags were tripped and an account would be challenged. It is based on the suspicion that it is an inauthentic or spam account, which is likely therefore to engage in inauthentic behaviour on the platform, perhaps in co-ordination with other accounts and-or as part of a network. If the challenge is not met by an account and it does not prove that it is being operated by a real individual and that person has not had accounts suspended previously, the account in question will be automatically suspended. There was a significant volume of such actions globally around 18 or 24 months ago. There is still a significant volume weekly but it is not as high as it was. I suggest that is evidence of our enhanced capacity to deploy those machine learning techniques to address this sort of behaviour.