Meta has made improvements to its artificial intelligence tool MetaAI but remains under scrutiny and needs to report back to the Irish Data Protection Authority (DPA) in October, the regulator said in a statement.
Meta said in March 2024 that it plans to train its Large Language Models (LLMs) using public data shared by adults on Facebook and Instagram across the EU. The tool was already launched in the US in September 2023.
Ever since, the tech giant has been under scrutiny from the Irish watchdog – Meta has its EU headquarters in Dublin – due to regulatory concerns related to the data protection safeguards of its AI tools.
After exchanges with the watchdog, and a pause on the roll-out due to “regulatory unpredictability”, Meta announced the launch of the tool across the EU in April.
The Irish DPA now said that Meta agreed to implement measures and improvements, including updated transparency notices to users, easier to use objection form, updated measures to protect users and updated risk assessments.
“As part of our ongoing monitoring, the DPA has required Meta to compile a report which, amongst other things, will set out an updated evaluation of the efficacy and appropriateness of the measures and safeguards it has introduced regarding the processing taking place. This report is expected in October 2025,” the statement said.
Last month, several national privacy watchdogs including those in Belgium, the Netherlands, and the German state of Hamburg, warned users of Meta’s Facebook and Instagram that they should object if they don’t want their public data to be used to train the company’s tool.
If users don’t object before May 27, Meta will automatically use their public data to train Meta AI.
Senior Meta officials have been critical of the EU rules and said that the launch in Europe took longer than expected.
