Reporting reveals Cambridge Analytica’s use of Facebook data

A voter-targeting model is an ordinary piece of machine learning: it is trained on records about people and then used to sort people it has not seen into groups worth different messages. Its accuracy depends almost entirely on how much it was given to learn from. That is why the question of where a campaign’s records came from is not a side issue.

Cambridge Analytica’s came, in part, from Facebook. The Observer and The New York Times reported on 17 March 2018 that the firm had obtained data drawn from as many as 50 million profiles, gathered through a personality-quiz app built by the academic Aleksandr Kogan, and had used it to build targeting models for the 2016 United States election and for Brexit campaigns. Christopher Wylie, a co-founder of the company, described the practice to both papers. Facebook suspended Cambridge Analytica and its parent SCL Group from the platform on 16 March, the day before the stories ran.

What followed was less a technical argument than a political one. Mark Zuckerberg was called before committees of the United States Congress and of the British Parliament. A #DeleteFacebook campaign ran for weeks. The lasting effect was on the standing assumption underneath most of the industry: that data collected for one purpose could be reused to train a model for another, and that this was a matter between a company and its developers. Ten weeks later the European Union’s General Data Protection Regulation took effect, written before the scandal but read afterwards in its light.

Corrections

• 2026-09-01 - This entry originally said Cambridge Analytica obtained data from up to 87 million Facebook profiles. The cited 17 March 2018 reporting said 50 million; 87 million was a Facebook estimate published in April 2018, after the events this entry describes. Corrected to 50 million throughout.