AI Governance Cannot Be Left for only the Giants to Negotiate

AI Governance Cannot Be Left for only the Giants to Negotiate

Last year, I had the privilege of moderating a Council of Europe Hackathon on democracy: Outsmart Disinformation, Protect Free Speech. The event sat within the wider New Democratic Pact for Europe, a project concerned with revitalising democracy at a time of rising polarisation, declining trust, digital manipulation and democratic backsliding.

Power is already being consolidated at extraordinary speed. The Stanford AI Index reports that nearly 90% of notable AI models in 2024 came from industry, up from 60% the year before. Frontier AI is increasingly dependent on vast compute, proprietary data, cloud infrastructure, specialist chips and capital expenditure at a scale only a handful of companies can sustain. In plain English, the future is being built in rooms most citizens will never enter.

This is why inclusive governance matters. If AI systems increasingly mediate public debate, employment, education, health, policing, journalism and political participation, then the rules governing those systems must be shaped by more than the people who build and profit from them. That should concern anyone who cares about democracy. Big tech is dominating the conversation to the point where democracies are left negotiating with systems they barely understand, cannot meaningfully inspect, and struggle to challenge.

At the Hackathon, the core democratic tension was clear: how do we counter disinformation without undermining freedom of expression? That same tension sits at the heart of AI governance:

• We need safeguards against manipulation, bias, deepfakes, algorithmic discrimination and synthetic propaganda.
• But we also need to preserve open debate, pluralism, dissent, artistic expression and civic participation.
• Heavy-handed control risks censorship. Light-touch optimism risks capture. AI governance must ask who benefits, who is harmed, who participates, who is excluded, and who has the right to challenge decisions.

Inclusion cannot be decorative. It must be structural.

• Young people need a voice because they will inherit the systems now being designed.
• Women, minorities and marginalised groups need a voice because digital harm is not distributed evenly.
• Journalists and civil society need a voice because they protect public scrutiny.
• Educators and cultural institutions need a voice because democratic resilience is built through literacy, imagination and trust, not just regulation.
• Local communities need a voice because the consequences of technological decisions are often felt closest to the ground.

The Reykjavík Principles for Democracy make this point plainly: democracy depends on participation, free expression, independent institutions, civil society, equality and meaningful public life. AI governance must be built around those principles, not added to them later as a compliance exercise. The danger is that AI governance becomes another elite bargain: governments seeking growth, companies seeking scale, and citizens being asked to trust the outcome.

That will not be good enough. Trust cannot be extracted from the public after decisions have been made. It has to be earned through transparency, participation, accountability and genuine contestability.

The future of AI will not be decided by code alone. It will be decided by who has power over the code, who can question it, who can refuse it, who can audit it, and who gets to imagine something better. If AI is going to shape democracy, then democracy must shape AI.

Mark Davis CEO, The Negotiation Initiative