Benjamin Harnett stared at the disciplinary notice. The numbers didn’t add up. One pull request per week. Twenty-five percent below some industry benchmark. The message was clear: his output fell short.
Harnett, a staff software engineer at The New York Times and chair of the Tech Guild’s generative AI committee, saw the fingerprints of two internal tools. DX tracked productivity metrics. Glean pulled from documents, emails and code comments. Together they painted a picture of his work that felt both incomplete and punitive.
That frustration boiled over last week. On May 27, 2026, the NewsGuild of New York filed two contract grievances and an unfair labor practice charge on behalf of the Tech Guild. The accusations cut to the heart of how companies can deploy artificial intelligence against the very employees who build it.
The union claims The Times used AI programs to surveil and monitor performance of unionized tech workers without notice. It failed to share critical information despite repeated requests. And it applied metrics in ways that violate privacy protections, job descriptions and bargaining obligations spelled out in the contract ratified in late 2024.
The contract was supposed to change the balance of power.
After more than two years of negotiations, a weeklong strike and an NLRB-certified union election in 2022, roughly 600 tech employees secured their first collective bargaining agreement. The deal delivered wage increases of up to 8.25 percent, just-cause protections against arbitrary firing, extra pay for on-call shifts and flexible hybrid schedules. (The New York Times, Dec. 11, 2024)
Yet the agreement also included language on performance monitoring and data use. Union leaders believed those clauses would prevent exactly this scenario. They were wrong. Or so they argue.
DX started as an internal tool pitched to improve developer experience. Managers spoke of company-wide measurements. Then the data turned personal. Benchmarks landed on individuals. Disciplinary meetings began citing weekly output figures. Harnett told The Verge that such metrics flatten nuanced engineering work. They ignore quality, problem-solving and the human judgment required to ship features that matter.
“All this data reasonably could be expected to help us understand how we’re doing,” Harnett said. “But not the way that they’re using it and implementing it, which we think is amounting to a de facto quota.”
Glean added another layer. The search tool queries internal wikis, GitHub repos, Google Docs and emails. Managers could, in theory, ask it for summaries of any employee’s contributions. Disciplinary notices started showing phrasing and structure that suggested AI generation. Harnett noted Glean’s tendency to hallucinate facts and send users on unproductive searches. Not exactly the foundation for fair evaluation.
The grievances accuse management of violating multiple contract sections. Protections around privacy. Rules on monitoring. Requirements to notify the union before introducing new performance systems. Obligations to bargain over changes that affect working conditions.
But the union’s complaint runs deeper than one contract. It filed an unfair labor practice charge because The Times refused three separate requests for information about AI use. The first came March 26. Others followed on April 22 and May 6. Federal law requires employers to provide data relevant to bargaining or contract enforcement. The Guild received silence.
The Times Guild, which represents editorial and other staff and is negotiating its own new contract, filed a parallel unfair labor practice charge over similar refusals. Susan DeCarava, president of the NewsGuild of New York, didn’t mince words. “Workers everywhere are under attack from the unethical use of artificial programs by bosses,” she said. “Sadly, New York Times management has proven themselves to be no different.” (NewsGuild of New York, May 27, 2026)
Harnett drove the point home in multiple interviews. “Using AI to surveil our work violates our contract and creates a skewed, inaccurate picture of our members’ work,” he told the New York Post. “Our work takes human judgment, problem-solving and skill that can’t be accurately assessed by AI analysis and proxy metrics. It’s the equivalent of setting an arbitrary story quota for journalists.”
The company pushed back. Spokesperson Danielle Rhoades Ha stated that The Times disagrees with the characterizations in the grievances. It will respond through the normal contractual process. The same holds for the information requests. The company has answered more than 80 such requests from the Guild in recent years. (The Verge)
This fight arrives at a charged moment. Tech workers built the systems now being turned on them. The Times Tech Guild stands as the largest union of its kind with bargaining rights in the United States. Its members include software engineers, data analysts, designers and product managers. They power the digital products that generate the company’s profits. And they understand AI better than most.
Harnett has stressed the union does not oppose artificial intelligence outright. Workers simply demand a voice in deployment. Blind metrics on tokens generated or pull requests merged create perverse incentives. They push quantity over quality. They distract from the actual goal: building good products.
The episode echoes broader tensions. Across industries, employers roll out automated tracking systems that score workers on opaque criteria. Unions have begun inserting language to limit such tools or at least force transparency. The Times case tests whether those clauses hold when management sees productivity slipping or costs rising.
Recent coverage shows the dispute escalated quickly. Only hours after the filings, outlets from Fox News to the New York Post picked up the union’s press release. Social media amplified the charges of surveillance. Yet few details have emerged about the precise algorithms inside DX or how managers query Glean. The company’s contractual response will likely reveal more. Or spark further litigation.
For now the grievances sit with an arbitrator. The unfair labor practice charge rests with the National Labor Relations Board. Both processes take time. Meanwhile, the Times Guild continues bargaining its own agreement. AI protections rank high on its list. The Tech Guild’s experience may shape what language ultimately survives.
And Harnett’s original concern lingers. When a machine reads your code commits, your comments, your draft specs and spits out a performance score, what exactly is being measured? Output? Efficiency? Loyalty? The union argues those questions belong at the bargaining table. Not inside a black-box dashboard.
The Times built its reputation on rigorous journalism that holds power to account. Its own tech staff now asks whether the company will apply the same scrutiny to the algorithms it deploys against them. The answer will matter far beyond the newsroom.
WebProNews is an iEntry Publication