In November 2023, the UK hosted the world’s first Artificial Intelligence (AI) safety summit where 28 states and the European Union agreed the Bletchley Declaration on AI safety.
The Declaration recognises the benefits and risks of AI technology, which poses a dilemma for governments across the world – how to strike a balance between harnessing AI technology, whilst guarding against malicious use. The First Minister, Vaughan Gething MS, has also recognised this balance as a relevant issue for Wales.
This article outlines how the Welsh Government and Senedd are responding to the risks and rewards of AI advancement.
Thinking like a human?
AI is a term generally used to refer to technology which can replicate human thought to carry out tasks. The phrase used to describe AI models at the cutting edge of this development is frontier AI, defined by the UK Government as:
… highly capable general-purpose AI models that can perform a wide variety of tasks and match or exceed the capabilities present in today’s most advanced models.
In its discussion paper on the capabilities and risks from frontier AI, the UK Government said that “progress in frontier AI in recent years has been rapid”. It highlighted the increasingly complex tasks that AI can undertake, such as:
… make new apps, score highly on school exams, generate convincing news articles […] summarise lengthy documents, amongst other capabilities.
The paper also suggests that progress is unlikely to falter due to the number of factors which drive improvement, including algorithms, investment and improvements in hardware.
Risks and benefits
The Welsh Government has identified both risks and benefits of AI for a range of issues in Wales. From public sector delivery to elections and the economy.
For example, the potential benefits of AI are recognised in sectors such as in teaching and the NHS and include specific innovations such as trials in cancer detection and diagnosis. However, the Welsh Government says it’s “taking a cautious approach” to the technology.
The Welsh Government issued a statement in February 2024 on steps taken to ensure that AI developments can process the Welsh language, including work with OpenAI to improve their chatbot GPT-4.
The Welsh Government has also spoken to the risks posed by AI, particularly in relation to election interference. Then Counsel General and Minister for the Constitution, Mick Antoniw MS, described AI as:
The biggest threat to our democratic system and structures at this moment in time…
Amendments tabled during stage two proceedings of the Elections and Elected Bodies (Wales) Bill sought to address this by improving transparency around when AI is used in the creation of Senedd election materials. Whilst they were not passed, the Counsel General recognised that they raised “significant issues”.
The Counsel General also confirmed that the UK had adopted UNESCO’s recommendation on the ethics AI, which provides a framework for AI development based on the “protection of human rights and dignity”.
Global action
Speaking on the need for AI regulation, the First Minister said in May 2024 that:
There are real threats as well as opportunities that come with this, and it's why I do think it's important that we work internationally […] Whether we want to or not, the fact that the rest of the world is looking at a regulatory approach to AI will matter to all of us.
World leaders recognised the need for international action at the first AI Safety Summit held at Bletchley Park on 1-2 November 2023, which culminated in the Bletchley Declaration. It focuses on states collectively:
- identifying AI safety risks of shared concern and building a shared scientific and evidence-based understanding of these risks; and
- building respective risk-based policies across countries to ensure safety in light of such risks, collaborating as appropriate while recognising approaches may differ based on national circumstances.
It was agreed that similar summits would be held every six months, with the latest one held in the Republic of Korea in May 2024. The next is due to be held in France later this year.
Regulatory frameworks
There is a patchwork of regulation across the world, with statutory and non-statutory approaches being taken by different governments. The drop-down sections below summarise some recent developments.
The UK Government set out a UK-wide approach on 6 February following its consultation,‘a pro-innovation regulatory framework for AI’. It establishes a framework, underpinned by the five principles of:
1. Safety, security and robustness;
2. Appropriate transparency and explainability;
3. Fairness;
4. Accountability and governance; and
5. Contestability and redress.
This is on a non-statutory basis, which the UK Government says offers “critical adaptability”. It will keep this decision under review.
The UK Government said it would continue “to assess any devolution impacts” as its policy develops and is “committed to engaging” the devolved governments on the design and delivery of the framework.
The UK Government also announced the creation of an AI Safety Institute, tasked with testing new types of frontier AI for potential harmful capabilities.
In 2024, the EU passed its Artificial Intelligence Act which it describes as the “first-ever comprehensive legal framework on AI worldwide”. The Act aims to provide clear requirements and obligations for AI developers and deployers on specific AI uses. It allocates different levels of risk to AI systems and bans those deemed high risk, such as social scoring - a system that ranks people based on social and economic activity.
In response to a written question on regulating AI in the workplace, the Cabinet Secretary for Economy, Energy and Welsh Language, Jeremy Miles MS, said:
AI regulation is a reserved matter. The Welsh Government, however, recognises that the EU’s Artificial Intelligence Act may impact businesses based in Wales delivering AI tools or services to the EU. We support the need for regulation, alongside standards, governance and assurance mechanisms.
There is currently no federal legislation regulating AI in the United States. However, President Biden issued an executive order (EO) on the safe, secure, and trustworthy development and use of AI. The EO includes measures requiring AI companies to share safety test results with the federal government, creates guidelines around privacy techniques and creates best practice on the appropriate role of AI in the justice system.
The UK and US have also signed a bilateral agreement on evaluating the safety of AI tools and systems.
Since the publication of its 2017 strategy on national AI development, China has brought forward targeted regulations around issues such as managing recommendation algorithms and deep fakes.
In March, Forbes reported that China has released a draft AI law, which appears to prioritise AI development.
It’s clear that the opportunity and risks of AI advancement apply as much to Wales as the rest of the world, and governments must respond to AI’s rapid evolution.
However, it seems fair to leave the final word on its regulation to AI itself. When asked how Copilot would conclude the article, it said:
In the symphony of algorithms and data, AI plays a hauntingly beautiful melody—a melody that resonates with both awe and caution.
Article by Madelaine Phillips, Senedd Research, Welsh Parliament
