The tech giant have said in a blog post that they have found a way to make its AI models better without training on its users’ data or copying it from their devices.

Instead, the company have outlined that they plan to compare a synthetic dataset to samples of recent emails or messages from users who have opted into its Device Analytics program using special devices.

The devices will be able to determine which synthetic inputs are closest to real samples, and according to the blog post, “only a signal indicating which of the variants is closest to the sampled data” will be sent to the company.

By Apply doing it this way, according to the company, it doesn’t access user data, and the data never leaves the device.

The findings from the plan will then be used to pick the most frequently picked fake sample to improve its AI text outputs, such as email summaries.