And from a legal and ethical perspective, like what does a fair relationship between creators and AI companies look like in your opinion? Like what does that, again, another big, big, probably unanswerable question, but how do we reconcile?

First of all, if you notice your work or you even suspect your work has been taken by a machine or discredited, that grievance should be taken very seriously.

Because what we've seen so far is that AI systems have a way of not even letting you know that your work was actually copied or used, right? So both on the legal and the ethical ground, we're seeing tons of cases, but courts are flooded with tons of cases around a lot of creative saying, well, I kind of suspect that is my idea.

That artwork, that looks like everything I've done, but the colors just shifted in a different dimension.

But there's something I want to distinguish very critically.

I think we have to separate two things that get blurred because the distinction matters very importantly.

So there is the input side, which is what goes into the training model of an AI tool.

And maybe where we can go next is like many conversations about AI focus on what the technology can do, but what do you think in kind of on the human side, like what do you think are the most important questions we should be asking about what AI is doing to human beings and creative or otherwise in publishing or otherwise, and maybe just broadly, like what are your thoughts?

What safeguards should be in place to ensure that AI doesn't further exacerbate these disparities? And can AI ever truly be fair in today's world?

In terms of safeguards against social inequality, we need to consider the necessity and the proportionality.

An AI tool might do really well, but if it's going to disenfranchise or remove certain workers from their jobs, then we need to ask ourselves, is it proportional to... getting us faster outputs? Do we want to sacrifice people's employment opportunities because we want to get faster output? Is it even necessary? Is this system actually needed at all? So those questions around necessity and proportionality is very important.

I think you've highlighted some work around developing a human rights AI impact assessment.

I think your office has championed privacy impact assessment.

These are not options.

When you are developing an AI tool, you have to look at the privacy impact assessments, the outputs of what you think this tool will do and take it through the entire lifecycle from design to deployment.

So, Jake, how can we better integrate and include diverse voices in policy conversations and raise the level of what you call race consciousness in AI governance?