AI for Everyone Framework

AI is improving quickly. Whether its benefits reach most of the world, or only a few places, depends on choices that governments and companies make now.

This framework sets out what ACATHRCA believes those choices should be. It is a living document: we will update it as the technology and the evidence change.

Affordable by design

Policy should reward efficiency and competition, so that the cost of capable AI keeps falling for people and small businesses.

Infrastructure everywhere

Connectivity, compute and local data centres should not be concentrated in a handful of countries.

Skills for all

Teachers, workers and public servants need practical AI training, not just awareness campaigns.

Requirements that scale with risk

• Developers of the most capable systems should test for serious risks and publish how they manage them.
• High-stakes uses, such as hiring, lending and healthcare, need human review and clear accountability.
• Everyday, low-risk uses should not face the same burden as frontier development.

Honesty with users

People should know when they are interacting with AI, and AI-generated media should be identifiable.

Companies should share what they learn about misuse, such as fraud, scams and cyberattacks, with each other and with authorities, so defences improve faster than attacks.

We will publish updates to this framework on our policy page, and we welcome feedback from governments, researchers and civil society.