In an age of distrust, decentralised AI could be the way to go

Closed frontier AI labs will always be vulnerable on trust grounds. Decisions are made by a handful of people, behind closed doors. Hugh’s April 2026 AFR column used Anthropic’s Australian visit as the news peg, then made a structural point: extraordinary growth does not solve a trust problem that is baked into concentrated control.

Bitcoin introduced the world to open-source, permissionless money and reduced the need for bank intermediation. The column asks whether a comparable approach — open models, verifiable training data, and architectures that do not require trusting a small number of firms — could ease public and political distrust of AI.

This is the digital-assets strand of the practice, written for a policy audience rather than a crypto audience. Read the original in the AFR.

Read the original: Australian Financial Review. Want to discuss the argument? Get in touch.