KPMG AI report contained hallucinated citations and statistics

Article excerpt

KPMG published a glossy report last year touting artificial intelligence's business benefits, complete with claims that AI could boost productivity by up to 26 percent and citations to specific research backing those numbers. There was one problem: the firm had used AI to write portions of the document, and the AI had fabricated most of it. The citations didn't exist. The statistics were algorithmic confabulation dressed up in the language of research. Financial Times analysis uncovered the contradiction at the heart of the embarrassment: a consulting giant had released a document arguing for AI's trustworthiness in enterprise settings while that same document proved the technology's fundamental unreliability. The irony cuts deep because KPMG is the kind of firm supposed to investigate and validate claims about emerging technology, not fall victim to the very hallucinations it was meant to be scrutinizing. The mistake reflects a broader pattern in corporate America right now: companies racing to deploy AI without adequate fact-checking, then publishing results that sound authoritative but rest on nothing.