Can AI Be Aligned with Democratic Values?

Carissa challenges assumptions about alignment, arguing machine learning cannot transparently enact justice and rule-of-law values.

Play episode from 21:50

chevron_right

Transcript

chevron_right

Transcript

Episode notes

AI ethics sounds reassuring. Carissa Véliz argues it’s a contradiction: you can’t build “ethical” systems on surveillance. The result isn’t accountability, it’s rule by opaque models you can’t inspect or appeal. That’s not ethics, that's the definition of Kafkaesque.

Joining Mark and Jeremy on Thinking on Paper, Carissa Véliz lays out three hard truths.

- Privacy is both a right and a duty. It allows lawyers, journalists, and citizens to act without intimidation. Remove it and democracy loses its working parts.

- Privacy is collective. The choices you make affect everyone else.

- Governments and Big Tech now co-produce surveillance; data moves in both directions, and history shows companies can rival states in coercive power.

Véliz also gets practical: what can be inferred from “just” location, why trading sovereignty for convenience breeds dependence, and how even small shifts — using Signal instead of WhatsApp, Proton instead of Gmail — can matter when 5–10 percent of people change their habits, the power shifts.

Surveillance isn’t the future. It’s the business model, and it works because we accept it.

Not on our watch.

Please enjoy the show.

And subscribe to help us keep telling the human story of technology.

Thanks,

Mark & Jeremy

Follow Carissa on X

Buy Privacy is Power--

TIMESTAMPS

(00:00) Trailer

(02:26) What Is Privacy

(05:31) Is Democracy At Risk?