Tech giants are enthusiastic about AI, but employees claim that the competition in AI is making their work more challenging.

      Employees at some of the largest AI companies report that the AI initiative is leading to longer hours and increased workloads.

      Tech firms have long claimed that AI would enable individuals to achieve more in less time. However, within several companies at the forefront of this technological advancement, some employees are highlighting a perplexing outcome: more efficient tools haven't necessarily resulted in shortened workdays.

      Workers from OpenAI, Anthropic, Meta, and Google shared with the BBC that demanding schedules and AI assignments are consuming their evenings and weekends. Interestingly, their employers have substantial evidence that AI can indeed enhance productivity.

      AI can increase worker output

      OpenAI has stated that Codex has become the primary AI tool utilized across all departments, including Legal, Finance, and Recruiting. As its functionalities improved, employees began assigning it longer and more intricate tasks, with many also using it to address responsibilities outside their standard roles.

      Anthropic's internal data supports this trend as well. By May, over 80% of the code merged into its repositories was written by Claude, while the average engineer was merging eight times more code daily in the second quarter of 2026 compared to 2024. Although Anthropic notes that code volume may exaggerate the actual productivity increase, they acknowledge that the growth still indicates a significant acceleration.

      Employees are not regaining that time

      A former technical employee at OpenAI recounted to the BBC that they worked approximately 70 hours per week, often utilizing weekends to catch up. OpenAI and Anthropic employees mentioned that intense release phases could lead to over 90 hours of work in a single week.

      At Meta, both a current and a former employee noted that workers were suddenly reassigned to urgent AI projects, resulting in some assignments spilling into nights and weekends. Former Google employee Amin Shali mentioned that AI priorities impacted him in a more indirect manner, linking overnight work to internal engineering issues as resources were shifted towards AI endeavors. He noted an improvement in his sleep and overall health after leaving Google.

      AI Chatbot

      Productivity increases may lead to higher workloads

      Ongoing research from UC Berkeley offers a possible explanation. During an eight-month study involving a 200-person tech company, researchers found that employees leveraging generative AI worked more efficiently and broadened their scope of tasks. Work began to encroach into previously designated break times, often without explicit instructions from managers for employees to extend their working hours.

      OpenAI has also suggested another potential outcome of increased efficiency. Its proposed Industrial Policy for the Intelligence Age advocates for trials of a 32-hour or four-day workweek as AI reduces routine workloads and operating expenses, with reclaimed hours potentially converting into permanently shorter workweeks or additional paid time off.

      This raises a significant question regarding AI productivity: if these tools indeed allow companies to reclaim hours of human labor, who benefits from those hours?