Sinéad Gibney (Dublin Rathdown, Social Democrats)
Link to this: Individually | In context

Today has been very interesting to draw together some of the very different strands that we all prioritise. I still have real major concerns about the regulatory framework itself and how it is going to operate and will have some questions about all of this and a little bit about the witnesses' council. In some ways, the council has been given an inherently impossible job to do. It is just going to be a question about the different ways the council will be able to grapple with that. In a response to a question from me, Dr. Scanlon said earlier that she is not a policymaker. I would argue that the council is a policymaking body. It has the potential to drive and influence policy. The witnesses have repeatedly said they do not have the answers and that they are prompting questions but we need those answers. We, as a society, need those answers. We, as policymakers, need those answers. I would encourage the council to ask the Government for what it needs, for example, if it needs data centre and environmental experts, and to really push for what it needs to be the body that right now rises to meet this challenge. Around the regulatory framework itself and what is coming down the track, I have concerns because of the time more than anything else. I formerly worked in IHREC, one of the bodies referred to. I would compare it in some ways to the national preventive mechanism, NPM, which is around the convention against torture. It actually has a similar model that the witnesses may like to explore.

My concern would be that the legislation has been in the pipeline for 12 years now. I do not see a regulatory framework designed by this AI Act that will happen in time to make the difference it needs to make. I have other concerns around the sandbox option and the fact, for example, that until you get that live data - real public data - what does the sandbox actually offer? Some of the things we touched on, for example, the algorithm and the failure to act around it and the sycophantic Open AI slip-up as it is being determined. What could prevent those things happening? There are a huge number of flaws in the responses we are getting. What can help? What can be better? What will allow us to stop these things from happening?