Flip or Flop? The Pacing Problems of Systemic Risk GPAI

Beskrivning

Drawing from socio-legal studies on technological governance and notions of the pacing problem between law and new technologies, Stefan Larsson presented ongoing work on aspects of how generative AI is regulated in the EU AI Act.

Specifically, the presentation focused i) the leap from the notion of "FLOPs" in the AI Act, that is, a fixed threshold of cumulative computation for training of a certain model, to the systematic risk level of the regulation. And secondly, ii) the presentation analysed the seemingly discretionary aspects of how this designation can be changed by the Commission, through various instruments and criteria pointed to in the AI Act.

That is, the two-tiered approach to general purpose Artificial intelligence (GPAI) in the AI Act means additional obligations to producers of ”systemic risk” GPAI — that can have “negative effects on public health, safety, public security, fundamental rights, or the society as a whole”. The designation to “systemic risk” is in the law linked to how much cumulative computation, measured in floating point operations per second (FLOPS), have been used in the training phase, or under certain criteria be decided by the Commission.

Furthermore, the Commission may change the threshold and can decide over what models with "high impact capabilities" should be designated as systemically risky by a rather discretionary set of criteria. Larsson analysed regulatory design in light of pacing problems and the legal unpredictability it creates, and speculated on possible implications of it.

Period	2024 nov. 5
Evenemangstitel	XXXIX NORDIC CONFERENCE ON LAW AND IT
Typ av evenemang	Konferens
Plats	Stockholm, SverigeVisa på karta
Omfattning	Internationell