Welcome to Euractiv’s Tech Brief, your weekly update on all things digital in the EU. You can subscribe to the newsletter here.
“‘Foundation model means an AI model trained on a large amount of data and capable to perform a wide range of distinctive tasks.”
-European Parliament’s co-rapporteur definition of the foundation model
Story of the week: Another intense week in the negotiations of the AI Act, with a series of compromises and non-papers floating around on some of the most sensitive parts of the text. On foundation models, the Spanish EU Council presidency put for the first time black-on-white the dispositions based on the tiered approach, which includes strong cross-references to the Copyright Directive and safeguards for rightsholders. Later in the week, the co-rapporteurs reacted by putting forth a first set of criteria for classifying foundation models as ‘high-impact’. The leading MEPs also introduced some additional notions of systemic risks, proposed moving the copyright dispositions in an article dedicated to generative AI and pushed to maintain their text on responsibilities across the value chain.
Things also started moving on to the other sensitive chapter: law enforcement. For the first time, an internal document of the EU Parliament considered dropping the total ban on real-time Remote Biometric Identification in exchange for some other bans, such as biometric categorisation and emotion recognition. The question is if it will be enough for left-to-centre lawmakers, while Brando Benifei told Euractiv this is just one o ..
Support the originator by clicking the read the rest link below.
