US Immigration and Customs Enforcement abandons machine learning to identify terrorists


In May 2018, US Immigration and Customs Enforcement abandoned the development of machine learning software intended to mine Facebook, Twitter, and the open Internet to identify terrorists. The software, announced in the summer of 2017, had been a key element of president Donald Trump's "extreme vetting" programme and expected to flag at least 10,000 people a year for investigation. ICE decided instead to opt for a contractor who could provide training, management, and human personnel to do the job. Concern remains that ICE will resume development on the software at a later date even though the ability of algorithms to scan social media and successfully predict crimes or terrorist attacks is at best unproven and at worst discriminatory and chilling.
tags: ICE, US, immigration, anti-terrorism, machine learning, algorithms
Writer: Drew Harwell and Nick Miroff
Publication: Washington Post
Publication date: 2018-05-17