SAM board members warn staff about budget language after AI bot mischaracterizes meeting discussions

SAM Board Members Question AI-Generated Meeting Coverage for Accuracy

Directors at Sewer Authority Mid-Coastside raised concerns about how an artificial intelligence bot summarizes their public meetings, citing mischaracterizations that shift the tone of discussions during a March 23 board meeting.

The board discussed Coastside News' "Coastie" AI bot, which automatically generates summaries of local government meetings. Director Paul Nagengast initiated the conversation after staff flagged language choices that could alter how readers understand what happened.

Nancy Marsh, a Granada Community Services District director who attends multiple meetings, submitted detailed feedback about the bot's output. She said the AI frequently mischaracterizes discussions in ways that feel editorialized, despite appearing neutral.

The Problem With Automated Summaries

Marsh identified three specific issues. First, the AI treats each meeting as isolated, missing how topics develop across multiple sessions. Second, it inflates language-describing straightforward questions as "pressing" and a third discussion of an ongoing issue as a "crisis."

Third, she said the bot's word choices introduce bias. "Although one might think the AI output avoids bias or editorializing, the frequent use of such characterizations is in fact editorializing and prone to mischaracterization of the human discussion," Marsh wrote in response to a reader poll.

Marsh proposed a practical fix: a reporter spending 20 minutes on a phone call to review the AI draft against what actually happened at the meeting. This human checkpoint would catch errors before publication without requiring full meeting coverage.

Why This Matters for Newsrooms

The complaint highlights a real tension in newsroom operations. AI summaries reduce labor costs and increase output speed, but they can distort meaning through word choice and context collapse. Local government reporting depends on accuracy-residents rely on meeting summaries to understand decisions affecting their water and sewer systems.

The SAM board represents three agencies serving about 24,000 people across Half Moon Bay, El Granada, Montara, and surrounding areas. Their meetings are public record, and how those meetings get reported shapes public understanding of local governance.

Marsh's suggestion-pairing AI drafts with brief human review-reflects a hybrid approach some newsrooms are testing. The cost of a 20-minute phone call is minimal compared to the reputational damage of persistent inaccuracy.