Artificial intelligence (AI) is growing rapidly in complexity and sophistication and is disrupting business processes around the world. Governments and regulatory bodies are struggling to develop frameworks which meet the ever-changing needs of the technology.
AI regulation in the UK
There is no AI-specific legislation in the UK. Rather, all UK businesses must take into account various existing legal obligations when developing and using AI, just as they would in adopting any other new technology. There are three main pieces of legislation which UK organisations need to consider:
Due to the digital transformation of financial service providers in recent years, AI has become increasingly crucial in the financial sector.
The regulators — the Prudential Regulation Authority (PRA) and the Financial Conduct Authority (FCA) — have not issued any AI specific guidance, but existing guidance such as the FCA’s Principles for Businesses (PRIN) and the PRA’s Fundamental Rules will be relevant to firms using AI. Under these rules:
The new EU rules
In April 2021, the European Commission published its draft Artificial Intelligence Regulation (draft AI regulation) introducing a first-of-its-kind, comprehensive, regulatory framework for AI. The rules are still in draft form and will be the subject of complexlegislative negotiations, so the regulation is not expected to be finalised and implemented before 2023. Some important aspects of the draft AI regulation are, briefly:
Application
It will apply to all organisations providing or using AI systems in the EU. It does not apply to personal, non-professional activity.
Restrictions
There is a “risk-based” model to categorising AI systems:
Potential fines
Fines of 10 million-30 million euros or 2-6% of the global annual turnover of the organisation, whichever is higher, may be levied. The level of fine imposed depends on the nature of the infringement.
The new EU rules: extra-territorial impact
The draft AI regulation will apply to organisations outside of the EU if they are: (1) “providers” (i.e., a developer) of AI systems for organisations in the EU; or (2) “users” (i.e., organisations who buy the AI system from the provider) of AI systems where the output produced by the system is used within the EU.
This means that both UK-based AI developers who sell their products into the EU, and UK organisations using AI technology which affects individuals in the EU, will need to comply with the new rules. The extra-territorial impact is significant and multinationals will need to consider whether to use a common set of AI systems compliant with the draft AI regulation (once finalised) or adopt different technology in different jurisdictions.
Will there be similar AI-specific rules in the UK?
The UK government has not indicated any plans to follow in the Commission’s footsteps, nor has it expressed a view on what the future of AI regulation will look like.
In the financial sector, the Bank of England and the FCA have launched the AI Public-Private Forum (AIPPF) to facilitate discussion regarding the use and impact of AI in financial services. A particular aim is to gather views on the possible future AI regulatoryframework, and in particular to look at how the UK can “strike the right balance between providing a framework that allows for certainty, regulatory effectiveness and transparency, as well as beneficial innovation”. The AIPPF have discussed and evaluated the Commission’s draft AI regulation, so it is clear that the Commission rules will be closely considered by UK legislators and regulators in developing any new framework. The AIPPF did not, however, give any indication of when a new regulatory framework will be draftedand implemented.
AI regulation in the UK
There is no AI-specific legislation in the UK. Rather, all UK businesses must take into account various existing legal obligations when developing and using AI, just as they would in adopting any other new technology. There are three main pieces of legislation which UK organisations need to consider:
Data Protection Act 2018 — contains a number of the provisions which are relevant to AI, for example, the obligation to explain to individuals any automated decisions being made about them and the logic behind those decisions. Equality Act 2010 — prohibits discrimination on the basis of “protected characteristics”, e.g., age, disability, race, religion. Firms must therefore take steps to avoid discrimination and bias in AI systems. Human Rights Act 1998 — requires that AI systems treat people fairly.
Due to the digital transformation of financial service providers in recent years, AI has become increasingly crucial in the financial sector.
The regulators — the Prudential Regulation Authority (PRA) and the Financial Conduct Authority (FCA) — have not issued any AI specific guidance, but existing guidance such as the FCA’s Principles for Businesses (PRIN) and the PRA’s Fundamental Rules will be relevant to firms using AI. Under these rules:
AI systems must have sufficient interpretability and ensure they are covered by risk management systems (see FCA PRIN 3 and PRA Rules 5 and 6).AI-assisted decisions must treat customers fairly (see FCA PRIN 6).Firms must be transparent and be able to explain AI decision making to their customers (see FCA PRIN 7).
The new EU rules
In April 2021, the European Commission published its draft Artificial Intelligence Regulation (draft AI regulation) introducing a first-of-its-kind, comprehensive, regulatory framework for AI. The rules are still in draft form and will be the subject of complexlegislative negotiations, so the regulation is not expected to be finalised and implemented before 2023. Some important aspects of the draft AI regulation are, briefly:
Application
It will apply to all organisations providing or using AI systems in the EU. It does not apply to personal, non-professional activity.
Restrictions
There is a “risk-based” model to categorising AI systems:
Unacceptable risk — AI systems which are a clear threat to the safety, livelihoods and rights of individuals (e.g., AI systems designed to manipulate human behaviours) are prohibited.High risk — there are 21 “high-risk” AI systems which are permitted, subject to various compliance requirements.Lower risk — there are certain lower-risk AI systems, e.g., chatbots, which are permitted but will be subject to transparency obligations.Minimal risk — all other AI systems can be developed and used without restriction.
Potential fines
Fines of 10 million-30 million euros or 2-6% of the global annual turnover of the organisation, whichever is higher, may be levied. The level of fine imposed depends on the nature of the infringement.
The new EU rules: extra-territorial impact
The draft AI regulation will apply to organisations outside of the EU if they are: (1) “providers” (i.e., a developer) of AI systems for organisations in the EU; or (2) “users” (i.e., organisations who buy the AI system from the provider) of AI systems where the output produced by the system is used within the EU.
This means that both UK-based AI developers who sell their products into the EU, and UK organisations using AI technology which affects individuals in the EU, will need to comply with the new rules. The extra-territorial impact is significant and multinationals will need to consider whether to use a common set of AI systems compliant with the draft AI regulation (once finalised) or adopt different technology in different jurisdictions.
Will there be similar AI-specific rules in the UK?
The UK government has not indicated any plans to follow in the Commission’s footsteps, nor has it expressed a view on what the future of AI regulation will look like.
In the financial sector, the Bank of England and the FCA have launched the AI Public-Private Forum (AIPPF) to facilitate discussion regarding the use and impact of AI in financial services. A particular aim is to gather views on the possible future AI regulatoryframework, and in particular to look at how the UK can “strike the right balance between providing a framework that allows for certainty, regulatory effectiveness and transparency, as well as beneficial innovation”. The AIPPF have discussed and evaluated the Commission’s draft AI regulation, so it is clear that the Commission rules will be closely considered by UK legislators and regulators in developing any new framework. The AIPPF did not, however, give any indication of when a new regulatory framework will be draftedand implemented.
Unique regulatory challenges
The use of AI brings with it a number of unique regulatory challenges. The new AI framework in the EU is a major step forward, but it remains to be seen how the draft framework will develop during the Commission negotiation period and also how the UK will respond.
The complexity of AI and how it affects consumers, particularly in financial services, certainly supports the position that AI-specific legislation is warranted in the UK.
