Meta dismisses leaks suggesting its AI scraped Australian news sites
The ABC, The Sydney Morning Herald, and some of News Corp’s assets were included in a list of sites that were allegedly scraped by Meta for the benefit of its AI models.
Meta has dismissed as “bogus” a whistleblower leak alleging it scraped several prominent Australian news sites, but stopped short of ruling out the possibility that the publications were used to train its AI models.
The purported leak accused Meta of targeting the ABC’s news site, along with News Corp’s news.com.au, the Daily Telegraph, Herald Sun, and The Australian; Nine Entertainment’s Sydney Morning Herald, and Seven West Media’s 7news.com.au, to train its AI models.
The list was published on 7 August by Drop Site News, a left-leaning Substack publication founded by former staff of the national security-focused site, The Intercept. The report contains a list of “roughly 100,000” web properties that were allegedly scraped to train Meta’s proprietary AI models.
The list was generated by a query run on the Meta database, using internal software called “Spidermate”, Drop Site said in its report.
