The Wrap: After a rocky 2025, newsrooms are expanding their AI programs
What the article is about
Published in The Wrap on January 7, 2026, this piece by Michael Calderone examines how major American newsrooms are extending their AI programs following a 2025 marked by public setbacks, staff pushback, and several high-profile failures. The article covers multiple organizations and draws a distinction between newsrooms that have adjusted their approach after mistakes and those still working out what AI should and should not do.
Context
The article opens with public opinion data that shapes the rest of its analysis: approximately 50% of Americans believe AI will have a negative impact on news. Past failures are the backdrop — AI-written articles with factual errors at Sports Illustrated and Gannett, an AI-generated analysis tool at the Los Angeles Times that generated internal controversy, and AI-generated podcasts at the Washington Post that met staff resistance before launch. These cases have raised the cost of further mistakes, yet newsroom leadership views continued experimentation as practically unavoidable.
How different newsrooms are approaching it
The New York Times maintains an eight-person AI team and draws a clear line: AI assists with research, summarization, and metadata, not article writing. The Times describes its editorial standards as non-negotiable in this context and has held to that position despite competitive pressure to move faster.
Business Insider has taken a different approach, launching an AI-generated news desk for quick-turnaround factual pieces. The organization distinguishes these AI-written briefs from its branded columnist content, which remains human-authored. The distinction is important to how Business Insider communicates with its audience.
The Washington Post’s AI-generated podcast rollout was marked by staff pushback. It proceeded, but the internal resistance was visible enough to warrant coverage.
Key argument
Calderone’s central point is organized around institutional integrity rather than technical capability. In journalism, audience trust is the asset. The newsrooms where AI is holding are the ones that defined in advance where human editorial judgment is not substitutable — and are maintaining that boundary. Organizations that deployed AI primarily for speed, without adequate editorial guardrails first, paid for it in credibility.
Who it is useful for
Editors, editorial directors, and journalists at news organizations of any size who are evaluating or revising their own AI programs. The comparative framing is useful: seeing what the Times, Business Insider, and the Post have actually done, rather than announced, provides a baseline for comparing against your own organization’s constraints and values.