Executives at major artificial intelligence companies have spent years predicting that AI tools will eventually cut down the amount of time people spend working, yet employees at some of these very firms say they are putting in far longer hours than a standard work week, according to accounts gathered by a national broadcaster. Workers at OpenAI, Anthropic, Meta and Google described weeks stretching to 70 or even 90 hours during intense project periods, alongside weekend shifts and abrupt reassignments onto AI initiatives.
An engineering director at Google claimed four years ago that AI would usher in a four-day work week by 2025. A year after that prediction failed to materialise, OpenAI went further, publicly urging other companies to trial a four-day week with unchanged pay, arguing that AI would soon handle enough human labour to make it possible. That proposal, detailed in OpenAI's earlier call for four-day workweek trials, framed AI as a tool that could reshape both workplace benefits and society more broadly.
Did OpenAI ever adopt the shorter week it recommended to others?
No — a former OpenAI technical employee who left the company last year said the firm never trialled the four-day work week it had publicly encouraged other businesses to test. Instead, the person described a demanding environment featuring regular "crisis meetings", weekend work and unforgiving performance reviews that could result in sudden departures.
"You go in on Saturday or Sunday just to catch up or make sure things aren't broken," the person said.
The former employee said they worked at least 70 hours a week at OpenAI, well beyond what they had experienced in earlier tech roles. Having since moved to an AI-focused start-up, they said their balance has since improved to roughly 50 to 60 hours a week outside of "sprints" — the industry term for concentrated stretches of long hours ahead of a product launch. At OpenAI and Anthropic, workers told the broadcaster such sprints can run for weeks and exceed 90 hours in a single seven-day period. Neither company responded to requests for comment.
What is happening at Meta with staff being "drafted" onto AI work?
At Meta, employees say they have this year been shifted without warning onto teams handling AI projects deemed urgent, a practice they call being "drafted" because there was no option to decline. One current and one former employee described the arrangement.
"They just move you over," the former employee said. "You can't say no – or if you do, you have to quit."
They said Meta has recently begun softening that rigid stance, though by that point many staff had already been pushed onto the work. Employees on these teams reported long hours stretching into the night and weekends, plus a sense of being "on call" even during off-hours. Meta's ongoing AI projects reportedly include a software-engineering tool and infrastructure designed to measure how well AI models replicate human tasks.
"You're literally working in teams of people trying to replicate humans doing jobs," the former employee said, adding that the work feels "endless".
A Meta spokesperson declined to comment. There is no legal limit in the United States on how many hours a worker over 16 can be required to work, unlike in the UK and Europe, where regulations cap a working week at 48 hours including overtime.
The pressure at Meta comes as the company has simultaneously pursued major job cuts alongside heavy AI spending. As previously reported, Meta cut 8,000 jobs while ramping AI investment to $135bn this year, its largest round of layoffs since 2023, underscoring how AI ambitions and workforce reductions have moved in tandem at the company.
Are workers outside AI teams also affected?
Yes — even staff not directly building AI tools say the technology has lengthened their hours. Amin Shali, a former Google employee, left the company in May 2026 after concluding AI was worsening his job, citing frequent overnight work caused by internal engineering failures that grew more common as Google diverted processing power and memory storage toward AI projects. A Google spokesperson declined to comment.
Since leaving, Shali said his sleep and overall health have improved, and that heavy dependence on AI tools within large tech firms fosters excessive pressure on engineers.
"AI is supposed to be doing so much for us now, so many more people should at least have better health and better sleep," Shali said.
What does the research say about AI and workloads?
A study conducted at UC Berkeley tracked hundreds of employees at a US tech company over eight months as they adopted AI tools, finding that workers "worked at a faster pace, took on a broader scope of tasks, and extended work into more hours of the day" — a finding also detailed in coverage from a report on the Harvard Business Review study. Researchers attributed part of the added burden to the need to constantly verify AI output.
Separate figures add further detail to this pattern. According to the 2026 Glean Work AI Index, workers using AI save roughly 11 hours a week but spend about 6.4 hours a week "botsitting" — monitoring and correcting AI systems — with 37% of AI interaction time going toward that supervision, 36% toward actually using the tools, and 27% toward learning new tools and building agents. A separate 2026 workplace survey from the Society for Human Resource Management found workers using AI save close to a full workday, or six hours, per week, but spend around four hours weekly fixing AI-generated errors. Meanwhile, an ActivTrak analysis of 443 million hours of work activity found the overall workday shrank by 2% while productive hours actually rose by 5%, suggesting AI is intensifying rather than shortening work.
Fortune reported in March 2026 that executives were using AI-driven productivity gains as justification to demand greater output from existing staff rather than shortening working hours, according to Fortune's reporting on AI productivity and workday length.
Why does the time saved by AI keep disappearing?
Neil Thompson, an innovation scholar at MIT, said it is unlikely that employees at leading AI companies are simply told to deploy time-saving tools and then stop working once tasks are complete.
"It leads to a situation where even if there were real time savings, it would be sucked up by the changes, and implementing them, and making sure they worked," Thompson said.
Thompson added that once a worker streamlines a process using AI, they or their employer tend to fill the freed-up time with additional tasks, whether to demonstrate value or because expectations simply expand.
"People assume that 20% less work means four-day weeks," Thompson said. "But new work emerges."
How does this connect to the wider AI jobs debate?
The tension between promised efficiency and mounting workloads sits alongside a broader debate over AI's effect on employment. Anthropic chief executive Dario Amodei has warned that as AI tools grow more productive, broader job losses could follow, while tech entrepreneur Charles Radclyffe has proposed a "minimum wage for robots" tax to manage the pace of automation and its impact on workers. Separately, ongoing coverage of whether AI will augment or replace jobs has highlighted early signs of AI's effects on job postings and employment patterns.
Key Facts
- OpenAI publicly urged companies to trial four-day work weeks but did not adopt one internally, according to a former employee.
- Workers at OpenAI and Anthropic reported sprint periods exceeding 90 hours in a single week.
- Meta employees describe being "drafted" onto urgent AI projects without the option to refuse.
- A UC Berkeley study found AI adoption led employees to work faster, take on more tasks, and work longer hours.
- Separate 2026 data from Glean, SHRM and ActivTrak show AI use both saves time and adds new supervisory burdens, with productive hours rising even as total workday length shifts only slightly.
