Meta has announced it is to begin training its artificial intelligence models with public content from across its platforms.
From tomorrow users of Facebook, Instagram, and Threads will start receiving notifications to explain the kind of data Meta will be using and for what purpose.
WhatsApp will not be included in the initiative, as messages sent and received on that messaging service are protected by end-to-end encryption.
From May, the social media company plans to begin harvesting public content such as posts and comments from its other platforms to train Meta AI models.
The company says the move will better support millions of people and businesses in the EU, by teaching its AI to better understand and reflect their cultures, languages and history.
People in the EU who use any of the platforms will start receiving notifications – in-app and via email – with details.
These communications will also include a link to a form where people can object to their data being used in this way.
Meta says it will honour “all objection forms we already received, as well as newly submitted ones”.
For those who don’t object, Meta says it will be able to train its AI models on relevant EU first-party user data from 27 May.
People’s private messages with friends and family and public data from the accounts of people in the EU under the age of 18 will not be used for training purposes.
Last month Meta AI launched in Europe.
It Is already using such methods in other regions to train its generative AI models, and follows a similar approach to AI training that Google and OpenAI use.
The announcement follows controversy over the use of copyrighted work to train Meta AI models.
Last month Irish authors joined forces to campaign against the alleged use of their copyrighted work to train Meta AI models.
A petition was launched by the Irish Writers Union after searches revealed that the work of prominent Irish authors, including President Michael D Higgins, Joseph O’Connor and Anne Enright, appear in a database used to develop the software.
In response Meta said “We respect third-party intellectual property rights and believe our use of information to train AI models is consistent with existing law.”
