It started with a handshake that never came. The union at The New York Times wanted to know how the paper uses artificial intelligence. Management said nothing. Just silence. So they sued.
The Tech Guild—around 700 engineers, designers, and data analysts—is ready for a fight. They say Times executives hid the rules. Or maybe the lack of rules is the problem. The Guild filed an unfair labor practice complaint this month. Why? Because the paper rolled out internal AI tools without bargaining. Or asking.
One tool is called DX.
On paper it’s a productivity helper. In reality it feels like a panopticon. Ben Harnett, an engineer there, says management first pitched it as a way to improve “developer experience.” Noble goal. Then came the change. The data stopped being about the company and started being about the person.
Benchmarks applied to individuals.
“Now people in disciplinary situations,” Harnett notes, “are suddenly having read back to them that they did only one pull request per week.” One. Against an industry average they don’t control. That metric says someone is 25 percent inefficient. Does it account for hard problems? Complex fixes? No. It flattens engineering into opaque numbers. Numbers that can end your career.
It amounts to a de facto quota.
The guild calls it surveillance. And it’s backed up. The other tool is Glean. It’s a search engine for everything. Wikis, GitHub docs, emails, Google Sheets. If you draft a feature description or leave a comment on a file Glean sees it.
Can your boss ask Glean “How is Ben doing today?”?
Apparently yes.
Harnett points out the disciplinary notices sent out lately have the rhythm and formatting of AI generation. Convenient. Glean isn’t perfect, either. It hallucinates. It sends people on wild goose chases through internal lore. But when it works it turns your own documentation into a witness against you.
Both the Tech Guild and the larger Times Guild —1,500 editors sales staff and support workers—filed complaints. The charge: Times refused to provide data on its AI use. A violation of labor law.
Times didn’t deny the tools. Spokesperson Danielle Rhoades Ha issued a statement. Standard corporate deflection. The paper disagrees with the “characterizations.” They’ll respond via “normal contractual process.” As if contracts matter when the ground shifts beneath your feet. They claim this is routine. They’ve answered over 80 similar requests recently. Maybe they are right. Or maybe volume is their shield.
Meanwhile the newsroom is negotiating a new deal. The stakes are higher. Journalists want AI transparency. If AI helps parse Epstein files or scan Gaza satellites great. Label it. Pay for it. Ensure a human is in the loop.
This isn’t unique to the Times. It’s industry-wide. ProPublica employees walked off for 24 hours last April over AI concerns. At McClatchy staff withheld bylines after their employer launched an auto-story writer. People are pushing back.
Harnett’s position is nuanced. Don’t ban AI. Just let us decide how it’s used. Right now metrics on token usage create pressure to do more. Not do better. Quantity over quality.
“It’s going to distract you from actually doing a good job,” he says. “Which is what the company should want.”
The problem is they don’t want that. They want data.
