Alexander Reese

Stable Resting Points for AI Governance

We often struggle in AI governance and ethics with continuous scales: When is a model fair enough to launch? When is it safe enough? At what point is a model too risky to be released in a certain way?

From what I've seen, the challenge is that without clear expectations, no position ever feels stable; every incremental move can be considered responsible. I believe that one of the greatest opportunities for AI governance is to provide principles with clear expectations to align on.

In negotiation, if someone says a price is $100 and "no dollar less," that sounds much more credible to me than if that same person had said, "Okay, I'll do $97 but no dollar less." By moving just three dollars, we move from a stable commitment to an unstable point where another concession just reduces the price but doesn't affect any principle.

What we need are principles strong enough to serve as resting points for clear expectations, but flexible enough to be revised over time as we learn.

← Back to Notes

About Alexander Reese

Alexander Reese is a Senior AI Ethicist within the Responsible AI pillar at Google, where he develops cross-functional strategies and rigorous testing methodologies to mitigate risks across generative models, AI agents, and the broader product ecosystem.

He holds a PhD in Economics, focusing his research on ethics through the lens of microeconomics and game theory. He earned a triple Master’s degree in Management with specialization in Economics spanning Germany, France, and the UK, alongside a separate Master’s degree in Philosophy.