Towards an Artificial Intelligence Governance Framework

Briefing prepared by Dr Joanna Nurse, June 2023, and endorsed by the Commonwealth Centre for Digital Health with the Platform for Planet, Place and People. 

Artificial Intelligence (AI) includes the perceiving, synthesizing, and inferring of information, demonstrated by machines, rather than humans or animals. AI is broadly characterized into 3 categories: Artificial Narrow Intelligence (ANI) which currently exists, with the future potential developments of Artificial General Intelligence (AGI) and Artificial Super Intelligence (ASI).

The Centre for AI Safety has released a statement saying “Mitigating the risk of extinction from AI should be a global priority alongside other societal scale risks such as pandemics and nuclear war.”  It is clear that the top tech leaders are worried. They believe AI is going too fast with the potential for negative impacts on society as well as humanity. The recent emergence of AI has raised a lot of questions including how powerful is AI? How can we control it? Or can we control it? Is it ethical? Does it represent an existential threat for humanity?

This briefing outlines the potential risks, governance options and the opportunities in order to maximise benefits of emerging technologies including AI, through enhanced capacity.

 Potential Risks from Artificial Intelligence: 

• Unemployment: with human jobs being taken over by AI
• Increased Inequalities: AI is already controlled by a handful of companies and algorithms that will further exaggerate existing human bias and inequalities 
• Quality of Education: the rapid validation of students original work is required as AI is being utilised to complete course work, and could reduce the value of educational certificates
• Misinformation: decision making by AI is not transparent and AI creates its own information that could destabilise governments and influence election outcomes and democracy
• Privacy and Human Rights: with facial recognition following individual activities remotely
• Cybersecurity: AI has enhanced potential for computer hacking that disrupts systems, essential societal and economic infrastructure
• Killer Robots: drones and robots that are able to kill humans already exist with the ability of drones to make autonomous decisions and to operate in swarms
• Weaponisation: AI could be utilized to create chemical weapons and create synthetic viruses with pandemic potential 
• Lack of Control: as AI advances there are concerns that it will become autonomous and humans will no longer be able to control it

Governance Options for Artificial Intelligence: international discussions are occurring in order to minimise current and future risks from emerging technology, including Artificial General Intelligence and Super AI, which could exist within 5-10 years. Risks are also posed by other forms of emerging technologies, including robotics, nanotechnology and the creation of synthetic infections. In the future a global governance framework needs to be developed, of which the following are a range of options that could be applied. Those in A could be developed unilaterally including as industrial standards and regulation, whilst those in B would involve global governance arrangements. These could be overseen by the creation of an international agency or platform responsible for addressing existential threats and risks for humanity, including AI and emerging technology.  
 

Governance Options for Artificial Intelligence and Emerging Technologies

A Ethical guidelines - for countries, professionals, education and the industry to follow, although these would be voluntary in nature, they could help to change the culture. Quality standards - that include either self-regulation or licensing companies which would have to register and could include independent monitoring and regulation. Inbuilt safety designs – for the intrinsic design of technologies to incorporate a no-harm to humans function, with the abilty of humans to over-ride controls.  Transparency of functioning – to monitor the design of technologies including decision making processes to avoid the generation of misinformation and disinformation.  Surveillance and monitoring – of risks from criminal and terrorist groups as well as non-intentional applications that could be harmful to humans.

B A Global Treaty – which would enhance legal powers but could take 5 years or more for countries to approve by countries; controlling non-state actors would still be a challenge. International Safety Protocols – such as that used for Nuclear Non-Proliferation with monitoring by an agency equivalent to the International Atomic Energy Agency. A Convention on the Weaponisation of Technologies – this could include the application of the Convention on the Prohibition for the use of Biological or Chemical Weapons; further updates to the convention could include other forms of AI weaponization.   Knowledge and Technology as a Global Good – to enhance equity and maximise benefits, as well as preventing the power of emerging technologies falling into the hands of dictators; this would ideally require a global governance infrastructure to apply ethical and safety standards and to facilitate the creation of benefits as global goods.

 
Enhance Capacity to Maximise Benefits: Digital solutions can be orientated for the benefit of humanity and identify strategic solutions to address our many and complex challenges, for example:

• Our Planetary Emergency: early identification of tipping points to prevent cascading climate change, scale solutions for renewable energy, energy efficiency and carbon capture.
• Education for Peace: open access resources to enhance evidence for peace, human rights and responsibilities, create global networks and share innovative solutions.
• Health as a Global Good: emerging technologies have the potential to transform pandemic prevention, health protection, disease treatment and the promotion of well-being for all.