Meta is making some minor concessions in its extremely dystopian plan to track employees’ mouse clicks and keystrokes in the name of AI training. The company has reportedly made some changes to the controversial project known internally as the Model Capability Initiative (MCI), according to a report in The Information.

Meta now plans to allow employees to “pause” the tracking for up to 30 minutes in the event they need to “check something personal,” the company told workers in a memo. A subset of employees will also be able to request to opt out of the program altogether, though this will be limited to remote workers with bandwidth concerns, people who deal with “sensitive” material and those who often work in spaces where they can’t easily keep laptops connected to a power source.

In other words, it sounds like the vast majority of Meta employees will still be required to allow their (nearly) every move to be tracked and recorded in the name of improving Meta’s AI models. However, the company did say that it had improved the software’s battery usage to address some employee complaints, Reuters reports. The company has faced protests from employees over MCI, which was announced last month just before the company laid off 8,000 workers and reshuffled thousands of others into AI-focused roles.

CEO Mark Zuckerberg recently defended the program to employees, telling them that “watching really smart people do things” is the best way for AI models to improve quickly. “The average intelligence of the people who are at this company is significantly higher than the average set of people that you can get to do tasks,” he said in leaked audio from a company-wide meeting last month.

“None of the data is being used for, like, looking at what people are doing, or surveillance, or performance track[ing], or anything like that. It’s purely just, like, we are using this to feed a very large amount of content into the AI model, so that way it can learn how smart people use computers to accomplish tasks. I think that this is going to be a very big advantage if we can do it.” He also added that if it works, “we’ll probably do more things like it” in the future.