Nowadays, it is difficult to do anything without touching “AI”. From the overnight sensation of ChatGPT to the catastrophic resignation of “The Godfather of AI” Jeffrey Hinton. AI is undoubtedly a “hot topic” globally. There is no sign of this changing any time soon.
In the midst of this AI-induced chaos, the UK government released a white paper on March 29, 2023, “An Innovation-Driven Approach to AI Regulation”. The white paper states that it is “Innovative, Proportional, Reliable, Adaptable, Clear and Collaborative”. In a white paper, he proposes a “framework” built around four key elements:
- Defining AI based on its unique characteristics to support regulatory coordination.
- Adopt a context-specific approach.
- Provides a cross-cutting set of principles to guide regulatory responses to AI risks and opportunities.and
- It delivers new core capabilities to support regulators, maximizes the benefits of an iterative approach, and ensures framework consistency.
These four key elements represent the framework’s aim to regulate AI without stifling innovation in the industry. The White Paper differs significantly from the Commission’s approach in both context and approach. In contrast to the specific and detailed requirements set out in the draft EU AI law (first published in April 2021 and updated in May 2023), this framework will be implemented by existing regulators. provides high-level guidance as proposed.
To guide you through the “regulatory fog,” our Technology, Data, Privacy and Cyber teams have set out five key takeaways.
1. Principles and laws
This framework is underpinned by a set of five principles used to guide the development and use of AI.
- Safe, Reliable, Robust – AI must be secure and risks must be identified, assessed and managed;
- moderate transparency explainability – Individuals or entities must be aware that AI is being used and be able to understand the decisions that AI makes;
- fairness – AI must not violate the legal rights of individuals or companies. For example, we must not discriminate or create unfair market results.
- Accountability and governance – The use and supply of AI must be supervised and clearly explained.and
- Objections and Remedies – If the AI makes a decision that is harmful or creates significant risk, we need a route to challenge that decision.
The reason there is no specific AI-focused legislation (at least for now) is to ensure that regulation is responsive to rapidly evolving technology, while at the same time ensuring that it is under the reassuring oversight of sectoral regulators. It’s about giving companies the flexibility to develop and use AI. These regulators are given the ability to exercise judgment when making decisions in their respective areas, allowing for an adaptive approach depending on the risks presented. Additionally, the agile framework is believed to avoid putting undue pressure on the market to comply with complex legislation, especially in relation to small businesses, including start-ups that lack the resources to comply. .
2. Innovation Over Risk – Unlocking Pandora’s Box?
The UK government has said it wants to harness AI and maximize its use, and has already identified AI as one of the five key technologies in the UK Science and Technology Framework. The white paper strongly emphasizes the importance of AI and specifically states how AI could impact the world as much as the invention of “electricity and the Internet.”
It should come as no surprise that the UK will take a market-friendly approach to AI regulation, resulting in longer execution times for legislative legislation to come into force. Nonetheless, it is surprising that the White Paper lacks details about the dangers of AI, and that such dangers could be adequately addressed by the White Paper’s proposals.
Meanwhile, on 11 May 2023 in Brussels, the Internal Market Commission and the Commission on Human Rights and Human Rights adopted a draft negotiating task on the EU AI law. The bill will need to be approved by the entire European Parliament before the European Council can negotiate the final form of the EU AI law (it will be held from 12-15 June 2023). In contrast to the UK framework, the law promises to set out detailed requirements for the supply and use of AI, which could pose greater challenges for innovators.
3. Context-specific analytics for AI
The white paper says the UK does not intend to assign blanket rules or risk levels to the use of AI across sectors or categories of technology. Instead, we choose a “context-specific” approach based on our evaluation of the results AI produces for a particular application. Furthermore, this white paper does not seek to apply any particular definition of AI. Instead, he defines the concept by referring to two characteristics: ‘adaptive’ and ‘autonomous’.
For example, the use of AI in “satellite broadcasting” to avoid traffic jams is likely to require low levels of regulation, whereas AI used to identify potential suspects in criminal investigations will require further regulation.
At first glance, this seems like a sensible and pragmatic approach, enabling low-risk AI applications without unduly burdensome regulatory obligations.
4. New AI Regulator Appears, Existing Regulator Appears
Instead of creating a new AI regulator, the whitepaper proposes that the principles be implemented by existing regulators (including the Information Commissioner’s Office, the Financial Conduct Office, and the Competitive Markets Surveillance Office). Existing regulators will be obliged to follow the Framework’s principles when assessing the use and supply of AI. However, application of the principles is subject to the discretion of regulators.
Existing regulators will have the power to promote clarity across AI regulations by issuing individual and joint guidance when AI spans multiple areas. However, there are few details about how this actually works.
Challenges arise when different regulators apply the five principles inconsistently or beyond their mandate, and when companies try to reconcile conflicting advice from multiple regulators. expected to occur. In addition, each regulator has its own enforcement powers, so two different companies may suffer different consequences for violating the same principles, simply because of their powers. While these challenges have yet to materialize, it is clear that regulatory cooperation and consistency will be key to the success of the UK approach.
Five. Enforcement is done with regulators
Luckily for companies (or unfortunately, depending on your worldview), the potential enforcement process for breaches of principle is still in its infancy, with government central risk departments providing broad support and oversight to regulators. It will be developed by regulators in conjunction with .
The decision to leave the enforcement and consequences of violations ambiguous may be cause for concern. The CMA’s recent decision to block Microsoft’s £55bn takeover of Activision Blizzard and the new powers the Digital Markets, Competition and Consumers Bill could give it (see our latest article here) , indicates a potential impact. Of all the things that regulators might affect the industry, this means they will have to keep an eye on further AI legislation in the UK with a cunning (A) eye.
Final thoughts (for now!)
While this may look like a ‘work in progress’, it is a clear marker placed on the international tech market to attract investment and point the way in Brussels. But arguably, the UK will have to follow the EU’s regulatory approach to a greater extent, and some question whether a stronger approach to addressing AI risks now would make the package even more attractive. right. Credibility depends on an early and proactive approach from regulators. Whatever the UK’s final regulatory position, cracking down on AI is imperative at a global level.
This whitepaper will be available for consultation (specifically for various questions contained in the whitepaper) until June 21, 2023.
