AI promises a four-day work week, but employees work 90 hours of overtime a week: the technology giants themselves have not achieved it. Technology giants such as Google, OpenAI, Anthropic and Meta invest hundreds of billions of dollars in developing AI every year. Executives have painted the same future for years: after AI takes over more and more jobs, humans can lose a few days.

AI promises a four-day workday, but employees work 90 hours of overtime per week: Tech giants themselves fail to deliver. Tech giants such as Google, OpenAI, Anthropic and Meta invest hundreds of billions of dollars in developing AI every year, and executives have painted the same future for years: after AI takes over more and more jobs, humans can work less days. OpenAI even publicly called on companies to trial a four-day work system without reducing wages, claiming that AI will greatly improve labor efficiency. However, a BBC investigation found that these companies that advocated shortening working hours themselves never fulfilled their promises. Former OpenAI employees revealed that they worked at least 70 hours a week during their tenure, and even more than 90 hours during the sprint period. They also came to work on weekends as usual to make up for progress or troubleshoot system failures. Ironically, the more AI improves productivity, the longer employees have to work. While Meta laid off about 10% of its employees this year, it also forcibly transferred a large number of people to the AI ​​project team. Employees privately called it "recruiting" and had no choice but to resign if they did not accept it. After joining the AI ​​team, it became normal to work overtime until late at night and on weekends, and be on call at any time even after get off work. Former Google employees said that after the company continued to shift processor and storage resources to AI projects, internal engineering system failures became more frequent, and they often had to stay up late to deal with problems. Their sleep and health conditions improved significantly after leaving the company. The time saved by AI is eaten up by new jobs. A study by the University of California, Berkeley, which tracked hundreds of employees at a technology company for eight months, revealed the core paradox: AI makes work faster and has a wider range of tasks, but the time spent on work in the day is actually lengthened. One important reason is that employees must constantly check whether the content generated by AI is correct. This "oversight cost" offsets the efficiency gains brought by AI. MIT innovation research scholars pointed out that even if AI really saves time, the saved part will be swallowed up by the work of system adjustment, implementation and ensuring normal operation. The deeper driving force comes from employees' own anxieties. Some people want to use the time they save to get more done, while others worry that they won't be able to prove their worth to their employer if they don't continue to improve their output. New jobs will always emerge, making the idea of ​​"reducing workload by 20% to a four-day work week" always remain theoretical. Anthropic CEO Amodei even bluntly said that the higher the productivity of AI, the more jobs may be eliminated in the future, which further aggravates the psychological pressure of employees who dare not relax. It stands to reason that AI has done a lot of things for humans, but the reality is that at least in the core areas of the technology industry, more people have not slept better and lived easier because of it. Instead, they have fallen into the trap of becoming more efficient and busier.