The Road to Responsible AI

The buzz around artificial intelligence (AI) is hard to miss these days. Debates regarding AI chatbots, voice clones and deepfakes are more prevalent than ever before. Generative AI has been defined as a type of foundation model that is “specially intended to generate, with varying levels of autonomy, content such as complex text, images, audio or video”. Foundation models are large, machine learning models trained on vast amounts of data and code that can be used to execute various tasks.

It comes as no surprise that governments are taking steps to address the multifaceted ethical and economic implications associated with this emerging technology.

European Union 

The EU's proposed AI Act centres around a classification system designed to categorise AI technologies according to the potential risks they may present to the health, safety and fundamental rights of end-users.

The AI Act classifies AI systems into four risk categories:

• Unacceptable risk: AI systems that pose a serious threat to fundamental rights or safety are prohibited.
• High risk: AI systems that pose a significant risk to fundamental rights or safety are subject to strict requirements, such as human oversight and transparency.
• Limited risk: AI systems that pose a limited risk to fundamental rights or safety are subject to less stringent requirements.
• Minimal risk: AI systems that pose a minimal risk to fundamental rights or safety are subject to the least stringent requirements.

Providers of high-risk AI systems must complete a conformity assessment prior to market entry. This assessment will involve demonstrating the system's alignment with the fundamental requirements outlined in the Act, which encompasses elements including data governance, transparency and human oversight.

European Union lawmakers reached a consensus to amend the EU’s initial AI Act draft, introducing a prohibition of biometric surveillance, as well as a requirement for developers of generative systems to disclose AI-generated content. The Parliament proposal will impose specific obligations on providers of generative AI systems. In particular, obligations focused on safeguarding against intellectual property rights infringement, and more specifically, copyright violations. For example, ensuring that the generative AI systems are engineered with effective safeguards to prevent content generation that violate such laws. In addition, these amendments also call for generative AI companies to provide summaries of copyrighted material extracted and incorporated into the training of each system, with an overarching view of increasing transparency as a whole.

United Kingdom 

The UK Government is currently developing a new AI regulatory framework of their own. The Government has articulated its intention of establishing a regulatory approach that fosters innovation while also upholding responsible and ethical AI deployment.

In March 2023, the UK Government published a white paper on AI regulation, which set out its proposals for the new framework. The white paper also advocates for a risk-dependent regulatory strategy, with varying conditions applicable to distinct AI systems. Notably, it states that AI systems categorised as high-risk, particularly those influencing critical life and livelihood decisions, would be subjected to the most stringent requirements.

The UK Government is currently reviewing feedback received from industry stakeholders in this respect.  

Recently, the Competition & Markets Authority (CMA) published their initial report on foundation models, underscoring their commitment to prioritising competition and consumer protection in the evaluation and expansion of these models. In their report they opt to address developer accountability to consumers by highlighting six main focus areas:

• Access: Ongoing ready access to key inputs.
• Diversity: Sustained diversity of business models, including both open and closed source models.
• Choice: Sufficient choice for businesses so they can decide how to us foundation models.
• Flexibility: The ability to switch or use multiple foundation models according to need.
• Fair Dealing: No anti-competitive conduct, including anti-competitive self-preferencing, tying or bundling.
• Transparency: Consumers and business are given information about the risks and limitation of this generated content in order for them to make informed decisions.

The UK is spearheading the global effort to develop AI guidelines. Their pro-innovation approach will undoubtedly be welcomed by the tech industry, considering the importance in protecting the public from potential harms.

Next Steps 

AI regulation in the EU and UK will continue to evolve, with a focus on balancing innovation and ethics. As these regulatory frameworks take shape, it must be noted that they will inevitably encounter ongoing challenges and complexities. AI technology is dynamic and constantly changing, meaning that unexpected issues may arise as it develops. Continuous monitoring, adaptation and collaboration with stakeholders will be crucial to address these emerging issues and ensure that the regulatory landscape remains effective and responsive in this rapidly advancing sector.

Gibraltar has a track record of being at the forefront of embracing digital technologies. In a similar vein to its pioneering approach in the crypto industry, Gibraltar is actively investing in understanding the intricacies of AI technology.