The newsroom temptation is obvious: let AI handle more tasks, save time, and redirect resources. But the risk is equally clear; publishing unverified, inaccurate, or tone–deaf content damages credibility faster than any efficiency gain can repair it.
Varun Shetty, Head of Media Partnerships at OpenAI emphasized this point in the recent AP Forward webinar:
“The most important thing is to make sure you have journalists and editors in the loop. If you’re using the technology to generate summaries, you want a human to review them.”
The pitfalls of skipping oversight have been on display. When Italian daily Il Foglio produced an edition written entirely by AI, its own editors acknowledged factual errors and a “flat” editorial voice. As one observer told The Washington Post, “It’s a good experiment to see the limitations.”
To guard against such missteps, organizations like the American Journalism Project are working with OpenAI to create local newsroom AI playbooks. These will focus on safety, editorial control, and audience transparency.
Actionable recommendations
- Enforce a human–in–loop mandate: no AI-generated copy should publish without editorial sign-off.
- Label AI involvement clearly: let readers know when and how AI assisted in content creation.
- Document failures: keep an internal record of AI output errors and the circumstances that caused them.
Questions for leaders
- How will our audience know they can still trust us in an AI-assisted newsroom?
- Have we defined the types of content that will never be generated by AI?
Related articles
