USI-founded fake news detection technology developer Fabula AI has been scooped up by Twitter, which will employ the spinout's talent to remove spam and abuse from its platform.

Fabula AI, a UK-based news forgery detection software creator spun out of Università della Svizzera italiana (USI), was acquired yesterday by social media network Twitter for an undisclosed sum.

Founded in 2018, Fabula AI operates a fake news detection platform powered by geometric deep learning algorithms that can handle datasets too large and complex for ordinary machine learning processes.

Fabula AI claims the product is 93% accurate in identifying fake news articles, according to TechCrunch. The approach centres on analysing flows of misinformation and their origins, rather than examining questionable content directly.

Aside from identifying fake news, physicists at USI have used Fabula’s algorithm to detect elusive neutrino particles known for holding neutral electric charge.

Twitter plans to assign Fabula’s team to a strategic research group tasked with building its proficiency in so-called graph deep learning. It will initially focus on cleaning up Twitter’s social feed, targeting adverse factors such as spam and abuse.

Fabula AI had planned to open up its technology for use in third-party applications later in 2019, however the fate of these plans following Twitter’s acquisition was unclear. The spinout does not appear to have disclosed equity funding.

Fabula’s geometric deep learning algorithms are the result of research led by Michael Bronstein, a former professor and researcher at USI’s Institute of Computational Science who is now chair in machine learning and pattern recognition at Imperial College London.

Bronstein was helped by researchers from institutions including EPFL, Nanyang Technological University, Singapore and Tel Aviv University.