Promise of less work with AI comes up against working hours of up to 90 hours; OpenAI did not test '4-day week'

Industry leaders say AI means fewer work hours — but their employees say they work up to 90 hours a week. Disclosure/BBC Brazil For years, executives from companies that invest hundreds of billions of dollars a year in the development of various artificial intelligence tools have insisted that this technology will ultimately make people spend less time working. A Google engineering director claimed four years ago that artificial intelligence would enable a four-day work week by 2025. Earlier this year, OpenAI took on that challenge. She formally encouraged companies to start testing the four-day work week (without changing pay), claiming that AI will soon be able to speed up human work so much that the corporate world should prepare. However, a former employee who left OpenAI last year told the BBC that the company never got around to testing the four-day work week he suggested others try while he was there. Instead, the person described what was often an exhausting work culture marked by frequent "crisis meetings," weekend work and "extremely rigorous" performance reviews that resulted in colleagues being suddenly fired. "You go there on Saturday or Sunday just to catch up on work or to make sure nothing is broken," the person said. Other companies have also pushed the idea that AI will effectively reduce the number of hours people need to work, for better or for worse. Anthropic boasts that its popular scheduling tool and chatbot Claude is capable of working alone for seven hours without interruption, which is equivalent to a full day of corporate work. Mark Zuckerberg of Meta said that artificial intelligence is beginning to fundamentally change the way people work and that AI tools can make employees so much more productive that a smaller workforce can accomplish what previously required a much larger number of people. Mark Zuckerberg testifies in landmark social media trial Although Meta laid off one in ten employees, Anthropic CEO Dario Amodei warned that as AI tools inevitably become more productive, this could mean even greater job losses. Despite these claims, workers at these same technology companies — who not only develop but also use the AI ??tools that are supposed to do at least some of people's work — claim to be working much longer than the typical five-day, 40-hour workweek. 'Sprints' Tech professionals in the US are generally well-paid, and while other industries — such as finance, law and medicine — also often require long work hours, a 14-hour day hasn't always been the norm in technology. For years, it was a more typical office job, 9am to 5pm. But the former OpenAI employee said he worked at least 70 hours a week, much more than in his previous tech jobs. This person now works at an AI-focused startup and said his work-life balance has improved, working about 50 to 60 hours a week, "outside of sprints." Sprint is a term used to describe a common period in technology companies in which people work long hours in the weeks leading up to the launch of a new feature or product. However, within AI companies and large technology companies that work intensively on AI projects, these sprints can be extreme. At OpenAI and Anthropic, for example, sprints can span several weeks and exceed 90 hours of work over a seven-day period, said technology professionals the BBC spoke to for this report. Neither company responded to the BBC's requests for comment. Artificial intelligence gains ground as a competitive differentiator in companies Artificial Intelligence 'Recruitment' At Meta, employees said they were abruptly allocated to teams this year to carry out AI work treated as urgent. They described it as "recruiting," according to one current and one former employee, because people don't have a choice. "They just transfer you," the former employee said. "You can't say no—or if you do, you have to quit." They said that although Meta had recently begun to rethink this rigid "take it or leave us" approach, by this point many workers had already been forced into this type of work. Work hours on these Meta teams are often long, with employees working late at night, on weekends, and feeling "on call" during the time they are not working, according to reports from current and former employees. In the US, there is no limit to the number of hours a person over the age of 16 can work. In the UK and Europe, for example, laws limit the working week to 48 hours, including overtime. Meta's current AI projects include creating an AI tool that can be used for software engineering and other types of work — as well as building an AI model infrastructure capable of measuring an AI model's success in replicating various tasks, the employees added. "You're literally working in teams of people trying to replicate humans doing tasks," the former employee said. This work, the person added, seems "never-ending." A Meta spokesperson declined to comment. Illustration shows a 3D printed Meta logo next to you