Code Dependent: Living in the Shadow of AI

Amsterdam built a list of teenagers its algorithm expected to turn criminal, and "the list was actually used by drug gangs to recruit 'easy targets'": the same file, protecting no one. That story is the method of the whole book. Instead of another Silicon Valley tour, the reporting goes to Nairobi, Salta, rural India, to the people who label the data and absorb the damage, moderators "who suffer from nightmares and depression due to their work." A reader who follows AI reporting closely still found "examples I've not seen before and direct interviews with people affected." The chapter most often named the strongest follows deepfake-pornography victims to a defense attorney who refuses the term digital assault: "the victim is not digital. They are human."

For a book by the FT's first dedicated AI editor, the vocabulary never gets sorted out: it "uses 'algorithm' and 'AI' interchangeably," and an AI-ethics academic writes that "the concept of 'algorithm' is never defined, despite being key to the text." The looseness runs down into the sourcing, assembled by "browsing papers from a few media-friendly AI personalities" rather than the research literature, which is why the specifics already feel dated two years on. And the ending. After 250 pages documenting how AI concentrates power in already-powerful hands, the proposed remedy is an AI treaty brokered through religious institutions, a fix one detractor calls "toothless and ornamental."