On 6 December, the last round of interinstitutional negotiations on the proposal for a regulation on artificial intelligence (AI Act) took place. However, due to the differences between the European Parliament and, the provisional political agreement was not reached.
On 7 December morning, negotiators had found a tentative agreement on foundation models, including a fundamental rights impact assessment for AI.
Whether to regulate foundation models had been a key obstruction heading into the talks, with major EU countries France and Germany pushing to give these models lighter-touch rules.
On Thursday, officials were still negotiating forbidden AI practices like using the technology for social scoring, and governance issues like the bodies that will oversee the Act’s implementation.
A key stumbling block is whether national security would be exempt from the Act’s scope, a request by EU governments that has previously been dismissed by members of the European Parliament. EU governments also oppose the Parliament’s plan to impose a full-on ban on facial recognition technology. One person involved in the room said that one main political group has been putting pressure on its members to align themselves with the governments’ position.
FEBIS is closely following the final negotiations.