Ars Technica newsroom AI policy

Ars Technica has published a formal newsroom policy that bars generative artificial intelligence from writing its stories, illustrations, or video. (arstechnica.com) The policy says Ars staff may use AI for transcription, summarization, brainstorming, research assistance, and administrative work, but “all of our reporting, analysis, and commentary is written by humans.” It also says humans make every editorial decision and remain responsible for accuracy, fairness, and ethics. (arstechnica.com) Ars published the reader-facing policy on April 22, 2026, after saying earlier that it would explain how its newsroom uses and does not use generative AI. The document frames the rules around two ideas: AI cannot replace human insight, and AI tools can still help professionals do better work when used carefully. (arstechnica.com) That puts Ars on the stricter end of a newsroom debate that is still unsettled across the industry. A February 2026 Center for News, Technology & Innovation briefing said many newsroom AI policies still stress transparency, human supervision, and verification, but often stop short of detailed operational rules. (cnti.org) Publishers are also trying to separate two different AI questions: whether reporters can use AI tools inside the newsroom, and whether outside AI companies should be allowed to ingest publisher content. Condé Nast, Ars Technica’s parent company, told the Senate in January 2024 that both AI training and AI output using publisher content should be licensed and compensated. (condenast.com) That business fight has accelerated in 2026 as publishers discuss marketplaces that would let AI companies pay for access to licensed articles and archives. Press Gazette reported on April 23 that publisher executives at its London event were openly discussing AI content marketplaces, bot blocking, and how much editorial work should be handed to machines. (pressgazette.co.uk) Industry groups are describing the same shift in more technical terms: less uncontrolled scraping, more structured licensing, tracking, and machine-readable delivery. WAN-IFRA wrote on March 24 that a “functioning AI content marketplace is beginning to take shape,” even though pricing and demand are still unsettled. (wan-ifra.org) Ars’ policy draws a bright line inside that larger scramble. The company is saying AI may help with the mechanics of journalism, but the journalism itself still has to come from people. (arstechnica.com)