Facebook Founder and CEO Mark Zuckerberg’s appearance in front of the European Parliament made an impact – either because of what he said or due to the awful mismanagement of the whole hearing. The fallout is still rumbling on over whether it was a coup that he attended or an embarrassment that he was allowed to get off so easy. Aside from the questions around data, fake news and marketing, he pulled out his “joker card” when answering probes around illegal content on the platform. The card in question is the statistic that Facebook’s AI can now “flag 99 percent of the ISIS and al-Qaeda related content that we end up taking down before any person in our community flags that for us.” Listeners are evidently supposed to be wowed by the anecdote. 99 percent is a lot. It’s almost everything. Except, it isn’t. First of all, does this number only refer to “ISIS and al-Qaeda content” – what about other illegal content? Secondly, 99% of what Facebook “ends up taking down” could mean anything. How much terrorist content flows freely on the platform, but isn’t flagged? Zuckerberg doesn’t say. Thirdly, “before any person in our community flags that for us” doesn’t mean no-one sees it. The question remains: how long does this content remain online before it is removed? This week research, provided by the Counter Extremism Project (CEP) and seen by New Europe, showed that it stays online for too long. The organization’s researchers found over a handful of illegal content that was available on Facebook and had been available for over a day up to a number of years. Much of this content are photos, news and videos promoting ISIS messages, and at times it includes footage of beheadings, amputations and shootings. So, either Facebook’s AI is not as effective as its CEO asserted or we have been deceived!