AI tool identifies extremist content online

The UK has launched an AI tool that detects online propaganda from terrorist organisations.

The internet is home to a whole host of people from different backgrounds and with different agendas. Not all of them are good, and so with a whole new world comes a whole new system of defence. The UK Home Office has worked in collaboration with ASI Data Science to create artificial intelligence capable of identifying online propaganda used by terrorist organisations to radicalise individuals. The AI algorithm can spot specific types of propaganda to remove from the internet. Platforms such as Google and Facebook already have AI systems to remove Isis propaganda but smaller companies may not have the resources to do the same. This new AI tool aims to allow all online platforms be free of Isis propaganda.

In order to train the tool, it was made to analyse thousands of Isis videos. It is able to distinguish between propaganda and news coverage of propaganda by detecting subtle signals in videos. Moreover, the algorithm can flag this content as people upload videos to the internet. It therefore prevents extremist content from reaching the internet. While the details of the inner workings of the algorithm are a secret, the Home Office has announced that it has an impressive success rate. The algorithm has an accuracy of 99.995 percent and can detect 94 percent of Isis video uploads.

Furthermore, the algorithm would flag on average 250 non-IS videos for review on a website with five million daily uploads. ASI Data Science points out that a analyst can easily manually screen this low number of false positives. This collaboration between the Home Office and ASI Data Science showcases AI and machine learning are helping to defend against terrorism online.