Mauritius Finance has published its latest quarterly magazine, showcasing a series of insightful thought leadership articles that explore the power of harnessing AI and blockchain and harnessing their potential for Mauritius as an IFC. The magazine highlights how these cutting-edge technologies can propel Mauritius into the digital age while strengthening its capacity for growth and innovation. Janesh Chuttoo was pleased to contribute to the publication. The magazine is available for download here with their permission.
__________________________________________________________________________________________________________________________________________
As artificial intelligence (‘AI‘) continues to evolve at a staggering speed, the need to establish a legal framework to nurture innovation while also balancing public safety and ethical considerations has never been greater.
Although AI is still in its infancy in Mauritius, the government has underlined its intention to harness the transformational powers of AI by coming up with a national strategy on AI back in 2018. As part of that strategy, the FSC came up with a Robotic and Artificial Intelligence Enabled Advisory Services licence and a bespoke rule to regulate that activity through the Financial Services (Robotic and Artificial Intelligence Enabled Advisory Services) Rules in 2021 (the ‘FSC AI Rules‘).
Why should we regulate AI?
Given its standing as an IFC of repute and also the increased role that AI will have in our daily lives, Mauritius should ensure that any AI system that is developed or otherwise used in Mauritius does not create risks that may lead to undesirable outcomes. For example, any legislation should ensure that the algorithm enabling an AI system to make decisions is capable of being audited by a human.
The current AI legal framework
The FSC AI Rules only apply to the extent that a licensee wishes to provide investment and portfolio management services through AI-enabled algorithms. While the FSC AI Rules are robust enough in addressing the governance aspects that a licensee should implement, the ability of the FSC to audit the computer program that takes the decisions is a desirable addition. Despite the existing FSC AI Rules, the unique challenges posed by AI – algorithmic bias, accountability and transparency, just to name a few – highlight the need for a comprehensive AI legislation.
Where should Mauritius start then?
The European approach to AI
It is widely accepted that the European Union Artificial Intelligence Act (the ‘EU AI Act‘) is, so far, considered the world’s most comprehensive legal framework to regulate AI across sectors. Mauritius should then look no further than the EU to draw its vision on the regulation of AI.
The EU AI Act defines 4 levels of risk for AI systems, namely: unacceptable risk, high risk, limited risk and minimal risk. For the purposes of this article, we shall only delve into unacceptable and high risks.
Unacceptable risk
AI systems that pose a clear threat to the safety, economic interest and fundamental rights of Europeans are prohibited. Eight practices are set out in that list:
- Harmful AI-based manipulation and use of deceptive techniques;
- Harmful discrimination based on certain vulnerabilities;
- Social scoring of individuals based on social behaviour or personal traits;
- Use of profiling to assess the risk of an individual committing a criminal offence;
- Using the internet or CCTV material in an untargeted manner to create or expand facial recognition databases;
- Use of emotional recognition patterns in workplaces or educational institutions;
- Biometric categorisation systems to infer sensitive personal data (e.g. race, political opinion, sexual orientation etc); and
- Real time and sweeping remote biometric identification in publicly accessible workspaces.
High risk
AI systems that pose serious risks to the health, safety or fundamental rights are classified as high-risk. Examples of such high-risk cases include AI components when used in:
- critical infrastructures (e.g. mass transport);
- educational institutions (e.g. access to education);
- employment and management of workforce (e.g. resume sorting for recruitment agencies);
- access to essential private and public services (e.g. credit scoring in banks);
- biometric identification, emotion recognition and biometric categorisation (e.g. identification of an offender);
- immigration, asylum and border control management; and
- administration of justice and democratic processes (e.g. preparation of court rulings).
High-risk AI systems are subject to a stringent system of controls before they are commissioned. Some of those controls include:
- risk assessment and mitigation systems;
- the testing and validation of datasets that are used in the AI model to ensure that they are error free, as far as possible;
- traceability of results;
- detailed documentation to enable authorities to audit the system;
- allowing for human oversight; and
- implementation of a quality management system.
Emphasis on transparency and accountability
The EU AI Act requires AI systems to have a certain level of transparency in them. In that regard, developers of AI systems must include specific disclosures to ensure that human users of the systems are informed that they are interacting with non-human intelligence. For instance, humans using chatbots should be made aware that they are interacting with an AI system.
To address one of the topical issues around deep fakes that has come to the fore with the widespread use of generative AI (e.g. Chat GPT and Deepseek), providers of such technology should ensure that AI-generated content is clearly and visibly labelled.
Governance
The European AI Office and authorities of the EU member states are responsible for implementing, supervising and enforcing the EU AI Act. Advice on and management of the governance aspects of the EU AI Act is carried out through the AI Board, the Scientific Panel and the Advisory Forum.
The EU AI Act is only part of a wider programme to support the development of safe and reliable AI, which also includes AI Factories, the Coordinated Plan on AI and an AI Innovation Package, from which Mauritius could draw inspiration.
Looking forward
AI is undeniably one of the most significant breakthroughs in human history. While the benefits from the adoption of such a system are immeasurable, if left unchecked it also has the power to create havoc on a global scale.
In its endeavour to ‘tame’ and foster the develop-ment of AI, Mauritius should not leave the regulation of AI to the last minute.
Partner