Google erodes user privacy to 'train AI models' in quiet policy update: report

Google quietly updated its privacy policy over the long July 4 holiday weekend to expand what it can do with user data, namely improve its artificial-intelligence abilities, according to tech blog Gizmodo.

The new policy replaced the word "language" with "AI" in a section referring to the "publicly available information" that Google uses to train its "models" for the benefit of users, Gizmodo says, citing the publicly recorded change log.

Google also added Cloud AI and Bard, a competitor to OpenAI's ChatGPT, to Google Translate as products that benefit from this training in the policy.

"This is an unusual clause for a privacy policy," said reporter Thomas Germain, who specializes in terms of service.

"Typically, these policies describe ways that a business uses the information that you post on the company’s own services," but Google seems to be giving itself permission to "harvest and harness data posted on any part of the public web, as if the whole internet is the company’s own AI playground," he wrote.

The policy change coincides with Twitter limiting how many tweets users can read daily, depending on their account status, which the Elon Musk-owned company partly justified to prevent "scraping people’s public Twitter data to build AI models."