Artificial intelligence has long been promoted by technology executives as a tool that could make work faster, more efficient and, eventually, less time-consuming.
But inside some of the companies building the world's most advanced AI systems, employees say the opposite is happening.
Workers at major technology and AI companies reportedly describe demanding schedules, weekend work and periods of intense pressure that can push their working hours far beyond the traditional 40-hour week.
Some employees interviewed by the BBC said they had worked as many as 90 hours in a single week during particularly intense periods.
The Promise of a Four-Day Work Week
Technology leaders have been predicting that artificial intelligence could dramatically reduce the amount of time people need to spend at work for several years.
In 2021, an engineering director at Google predicted that AI could eventually help make a four-day working week possible.
OpenAI later encouraged businesses to consider experimenting with a four-day week while maintaining employees' existing pay.
The argument is that increasingly capable AI systems could automate repetitive tasks and allow employees to accomplish substantially more within fewer hours.
However, a former OpenAI technical employee who left the company last year told the BBC that the company had not actually tested the four-day work week during their time there.
Instead, the former employee described a demanding environment involving frequent crisis meetings, weekend work and highly competitive performance assessments.
The employee said there were occasions when workers would return to the office on Saturdays or Sundays simply to catch up or make sure systems were functioning properly.
AI Companies Are Working at High Speed
The pressure is not limited to OpenAI.
Anthropic has highlighted the ability of its Claude AI system to work autonomously for extended periods, including several hours at a time.
Meta CEO Mark Zuckerberg has also predicted that AI will significantly transform how companies operate and allow smaller teams to accomplish more.
At the same time, Meta has announced significant job cuts, while Anthropic CEO Dario Amodei has warned that increasingly capable AI systems could eventually result in substantial employment losses.
This creates an apparent contradiction.
AI is being promoted as a technology that can reduce the amount of human labour required, yet employees developing and deploying these systems say they are often working exceptionally long hours themselves.
Long ‘Sprint’ Periods
Long working hours have existed in the technology industry for years, particularly around major product launches.
Technology companies commonly refer to these periods as “sprints”, when employees work intensively to complete a product, feature or major release.
However, workers at some AI companies say these periods can last considerably longer than traditional technology-industry sprints.
A former OpenAI employee told the BBC that they regularly worked around 70 hours per week while at the company.
The person has since moved to an AI-focused start-up where their typical workload is closer to 50 to 60 hours a week, outside particularly demanding periods.
Workers familiar with AI development at OpenAI and Anthropic told the BBC that some intense work periods can continue for several weeks and, in extreme cases, reach more than 90 hours in a seven-day period.
Neither OpenAI nor Anthropic provided a response to the BBC's questions about the claims.
Employees Say They Are Being ‘Drafted’ Into AI Teams
Workers at Meta described another source of pressure.
Current and former employees told the BBC that some staff had been moved onto urgent AI projects without being given much choice in the matter.
Some workers reportedly referred to the process as being “drafted”.
One former employee said workers were effectively told to move to the new teams and that refusing could mean leaving the company.
Meta later began easing that approach, but employees said some workers had already been reassigned by that point.
The workloads on these AI teams were reportedly intense, with employees working late into the evening and during weekends.
Some workers also described feeling as though they remained effectively on call outside their normal working hours.
Different Rules Around Working Hours
Working-hour regulations differ considerably between countries.
In the United States, federal law generally does not impose a maximum number of hours that employees aged 16 and over can work, although overtime rules can apply to eligible workers.
In the UK and parts of Europe, regulations generally limit the average working week to 48 hours, including overtime, although workers can sometimes agree to opt out of certain limits.
The difference highlights how working cultures around AI can vary significantly depending on where employees are based.
The AI Workload Is Not Limited to AI Engineers
The pressure associated with the AI boom can also affect employees whose primary jobs are not directly related to developing AI models.
A former Google employee, Amin Shali, said he left the company in May after becoming frustrated with the way AI-related priorities were affecting his work.
He claimed that internal engineering systems experienced problems more frequently as Google shifted important computing resources, including processing capacity and memory, towards AI projects.
According to his account, resolving those issues sometimes required employees to work late into the night.
The Paradox of AI-Powered Work
The experiences described by workers raise an important question about the future of employment.
If AI can eventually complete tasks that previously required several hours of human labour, will employees actually gain more free time?
Or will companies simply use the additional productivity to increase expectations and require workers to produce even more?
The answer may depend less on the capabilities of AI itself and more on how businesses choose to use those capabilities.
AI could potentially shorten working hours, increase productivity and allow employees to focus on more meaningful tasks.
But if companies use AI primarily to accelerate deadlines, increase output and reduce staffing levels, workers could instead face greater pressure.
For employees currently building the technology, the promised four-day work week may still feel a long way away.
