Preamble

The governments of the world have failed to assume their responsibility to protect their peoples from the existential and other threats posed by artificial intelligence. Commercial and geopolitical competition are driving out the international cooperation needed to keep us all safe from AI’s worst risks.

As a result, one of the most disruptive societal and economic transformations of the century is being shaped by a small number of mostly private companies, concentrated in two countries, with insufficient oversight and accountability. 

Some of these companies are crying out for stronger governance of their own work. Leaders should listen to them. The USA and China must take the lead in initiating the negotiation of international governance to establish common AI safety standards.

A more principled approach is overdue; one that puts people ahead of profit and power, is more globally inclusive, and takes a longer-term view. 

AI is not a force of nature or a destiny. It is built by companies and sold in markets. Like every other technology, it must be subject to the law. Some of its promises are real, but so are its harms, which may fall hardest on those with the least power to decide its future.

Leaders have collectively chosen to restrict nuclear weapons, ban chemical weapons, and repair the ozone layer. As former Heads of State and Government, we are acutely aware of the challenges faced in bringing about these achievements. But we also know that difficult things can be achieved with decisive leadership. 

Responsible leadership on AI is long overdue, and needed now with ever-increasing urgency. We therefore call on leaders to manage AI based on the following principles.

The Principles

1. Humanity

AI must be governed to serve human beings, not to threaten their human rights or their very existence. AI must never control humans. 

2. Safety

A government’s first responsibility is to protect its citizens. Leaders must act to keep their people safe from the risks AI presents. They cannot leave this responsibility to the companies developing AI.

3. Cooperation

No country can manage the risks of AI on its own. AI's capabilities cross national boundaries. International agreements can encourage responsible development while allowing innovation and commercial competition. The goal of international cooperation on AI is not to slow progress, but to ensure that AI is safe, trustworthy, and beneficial for everyone.

4. Trust

Leaders must tell their people the truth about AI: what it can and cannot do, what dependencies it may create, and how any gains or losses are likely to be distributed. They must share what is known about its harms and risks, and what is not yet known by anyone, including its makers. They must act in their entire country’s interests, not those of a small group of people.

5. Accountability

Leadership means governing – enforcing the law that exists, writing the law that is missing, and breaking concentrated power where it has grown excessive. Setting rules does not mean falling behind. Those who develop AI must be held to account for their products on behalf of all people.

6. Independence

A leader cannot govern an industry while depending on it – for advice, for infrastructure, for campaign support, for the promise of jobs. Core functions of the state, from public services to essential infrastructure, must not be outsourced to systems that neither the public nor the state can examine or hold accountable.

7. Restraint

The legitimate use of force is the gravest power a leader holds. It must remain a human responsibility. Leadership means rejecting the use of weapons that select their own targets, and commercial AI systems that shape who is targeted and how.  

8. Solidarity

Leaders must create opportunities for all those affected by AI to be involved in decisions about it, not merely to consume it or suffer from it on terms set elsewhere. This includes the communities bearing AI's environmental costs, the countries excluded from its benefits, and the workers whose livelihoods, healthcare, and rights are now at risk. 

9. Integrity

Leadership means holding the state's use of AI – in policing, in welfare, at borders, in war – to human rights standards. A leader cannot demand of companies what their own government will not practise.

10. Long-term thinking

The consequences of today's AI decisions will outlast every current term of office. Leaders must act as stewards for the young and future generations. Leaders must weigh the promise of short-term gains against the risks to the long-term health and safety of their societies and economies.