Meta Shifts the Blame + Do Data Center Bans Work? + The Final HatGPT

The night before recording, Kevin Roose spilled a glass of wine on himself at dinner with a "very fancy technology person" and tried to pass it off as excitement about media's future. That anecdote set the tone for an episode of Hard Fork that moved from a historic legal settlement against Meta to a provocative argument about why blocking data centers won't meaningfully slow AI, and finally to the bittersweet retirement of the show's long-running "Hat GPT" segment.

Meta's $17 Billion Settlement: The Terms and the Spin

The dominant news of the week was Meta's agreement to settle a multi-state lawsuit brought by 47 states, D.C., and several U.S. territories for up to $17.1 billion—the largest settlement in the company's history. The case centered on two core violations. First, Meta knowingly collected data from millions of children under 13 without parental consent, a direct breach of federal privacy law. An unredacted version of the lawsuit provided "tons and tons of evidence" that the company knew underage users were on Instagram and Facebook but did nothing. Second, the attorneys general argued that Meta's product features—push notifications, ranking algorithms, and the lack of enforced screen-time limits—were designed to create addiction in teenagers.