Artificial Intelligence in EU Investment Funds

1. Introduction

A Trends, Risk and Vulnerabilities report from the European Securities and Markets Authority (“ESMA”) released last week (the “Report”) on artificial intelligence in EU investment funds shows that the investment opportunity and operational efficiencies of AI in investment funds are being active considered and trialled by asset managers, with interesting results over the last few years. 

The report examined the use of AI by investment funds domiciled in the EU both from the perspective of (i) use of AI tools by asset managers in making investment portfolio decisions; and (ii) investment in companies producing AI products and services by investment funds domiciled in EU member states. 

With the introduction of the AI Act in August 2024, the EU has been ahead of other regulators globally in creating a regulatory framework for AI. As has been the case over the last few years, the EU regulatory bodies have been proactively and collegiately working with EU member states and economic stakeholders to create workable, proportionate and responsive regulation for financial services and products being offered to consumers and other investors resident in the EU.

2. Key Takeaways / Executive Summary

2.1 While there is a clear mandate to create a regulatory framework for use of AI by investment funds in the EU, this will need to take account of systemic risk and financial vulnerabilities to ensure the EU financial eco-system remains shock-proof and robust. The limited number of companies offering AI tools and services may pose a potentially systemic risk, noted as a growing concern for the Financial Stability Board.

2.2 The AI investment landscape is competitive with 86 fund managers offering funds domiciled in 12 EU member states, most of which cater to retail investors, with just over €13bn AUM, representing 0.1% of AUM of all UCITS in the EU;

2.3 Although there is clear appetite from investment houses to harness the processing efficiencies, with related cost-savings, that natural language processing AI and generative AI can bring in assessing portfolio holdings and managing risk and compliance, investment houses are yet to rely on AI. On the other side of the fence, Opportunist investors who were initially interested in investing in AI do not seem to have translated into lasting investment, with investment in funds promoting their use of AI in the investment process having declined since 2023;

2.4 Funds which used AI systematically (i.e., where an AI-based algorithm primarily determined investment strategy) remained fairly consistent in value and number whereas funds which used AI as a part of its investment strategy (non-systemically) grew substantially in value and number;

2.5 Given the proactive approach from the EU regulators to date, industry expectation in Ireland is that the EU AI Act and implementing measures will develop and evolve to meet this emerging market sector and indeed continual review is envisaged in the AI Act with provisions for an annual review by the European Commission of types of high-risk AI systems and the prohibited AI practices, and a wider review every four years on amendments needed to extend the scope of the act and enhancements of the supervision and governance system. 

3. ESMA Report Findings

3.1 Use of AI by Asset Managers - What’s in it for us?

The backdrop of increasing competition in the investment management industry has created a strong appetite for time- and cost-saving measures. The Report notes an increase in patent filings for asset allocation and portfolio management machine learning AI which can enhance the efficiency of cash flow and liquidity management, automate asset class rebalancing, improve valuation and forecasting methods and interpret unstructured and alternative data.1 

However, this has yet to translate into adoption of these AI tools at a high level in the decision-making process. It seems that like other sector stakeholders, asset managers are testing and watching the development of AI tools but are not yet ready to trust the tech to make the big decisions. To date, AI tools are being predominantly used to inform and support human decision-making, to enhance administrative tasks and to monitor risk management and compliance.

Funds which have as their investment strategy investment in AI-driven companies have seen an increase in investment of over 50% since 2023. The market value of these positions has reportedly almost doubled, as the underlying companies capitalised on their market positions. 

Walking the Walk - Performance Analysis

In a large-scale analysis of investor documentation of 44,000 EU investment funds in 22 languages over the past 4 years, the Report found that of funds which disclosed integrated AI in their investment process:

• The majority of the 145 funds that indicated use of AI in their investment strategies did not explicitly advertise their use of AI, and none included use of AI tools in their regulatory documents;

• These 145 funds were spread among equity funds (74), mixed-assets (31), alternative assets (29) and fixed income (10) investment mandates; four of these were ETFs;

• Apart from two index-tracking ETFs, these 145 funds were actively managed;

• The landscape was competitive with 86 fund managers offering funds domiciled in 12 EU member states

• Most of the funds (133) catered to retail investors;

• The funds had just over €13bn AUM, representing 0.1% of AUM of all UCITS in the EU;

• Funds which used AI systematically (i.e., where an AI-based algorithm primarily determined investment strategy) remained fairly consistent in value and number;

• Funds which used AI as a part of its investment strategy (non-systemically) grew substantially in value and number;

• 30% of the funds used AI as a key role in investment decisions, which funds tended to be smaller portfolios representing 5% of the testing sample’s AUM;

• There were net outflows of investors in funds using AI over the period, more so than for other EU investment funds by reference to respective AUM;

Although a third of these funds used ‘AI’ or ‘ML’ in their name, less than half of that third used AI to primarily determine investment strategy. While this may pose a concern for ‘AI-washing’ of funds, it is not seen as a regulatory concern and was recognised as asset managers marketing such funds to investors with appetites for innovative and sophisticated investment mandates. 

The perceived savings in time-costs in using AI to gather and process documentation to support investment decisions is likely, in the near term, to be outweighed by the investment cost in such AI tool. While equity funds using AI offered lower operational costs for investors, fixed-income, mixed-asset and alternative-asset funds using AI did not. Notably, there was no indication that use of AI created an increased cost for investors. 

There was no evidence to show that funds have achieved higher yields by using AI in the investment process than not: since 2022, ESMA found that funds with a stated use of AI did not have a significant advantage over non-AI funds in relation to the alpha between average returns and risk-adjusted returns. This remained consistent when the sample was controlled for size, age and geographic focus.

The Report notes similar findings by studies of US mutual funds: a study of US mutual funds by Praxmarer and Simon (2024) found that AI-managed funds performed similarly to human-managed funds. Interestingly, a study by Chen and Ren (2022) found that while AI-powered US mutual funds significantly outperformed human-managed US mutual funds, they did not outperform the market.

The Report also notes that asset managers’ testing and experimenting with AI is dependent on budget and IT savvy users and consequently larger asset managers can afford to experiment and trial more than smaller asset managers, and indeed larger asset managers have taken over tech start-ups to acquire talent and tech. With more limited budgets, smaller asset managers tend to use customisable tools, which have related risks of dependency on external expertise and service disruptions with a lack of control over the tech. 

3.2 Portfolio Investment in AI 

As AI has generated interest in sectors such as healthcare, manufacturing and finance with much anticipated revolutionising of these sectors, so has the opportunistic interests for investors in companies creating robotics, automation and big data. 

Benchmarking

AI indices have been created by various ratings agencies which, similar to IT indices, are highly concentrated, given the small number of entities in the sector. Nasdaq, WisdomTree, VettaFi, Solactive, Morninstar, Indexx and S&P have each created indices of up to 84 entities, with levels of concentration of the top 10 constituents between 23% to 60%. 

Unsurprisingly, the so-called Magnificent Seven IT companies2 appear in the AI indices. However, there is little other commonality of constituents among the seven AI indices with just 16 of the 198 total constituents included in all 7 indices. More than half of the firms appear in only one of the seven indices and just over 40% are common to two or more AI indices. Of these 40%, just 7% are EU domiciled – the remaining 75% are US, 12% are East Asian and 5% are Israeli.

The diversity in selection criteria for these indices is unsurprising given the lack of codified categorisation of firms as AI / IT / FinTech and so forth. The indices include companies engaged in the AI value chain such as developers, enablers, engagers and enhancers. Also reflective of the novelty of AI is the lack of clarity around what ‘AI’ is – does it include hardware, software, AI-adjacent or a combination of the above? 

The Report notes the divergence of these seven indices from the S&P 500 IT index and from each other since the release of AI in 2023/4, and also the underperformance of the AI indices relative to the S&P 500 IT index. Prior to then, the seven AI indices were fairly closely correlated but after 2023, they have diverged significantly and, as already noted, have differing criteria for constituent companies. 

Fund Performance

Assessing the AI indices by focusing on those constituents that appear in two or more indices, which includes mid-, large- and mega-cap companies, and then cross-referencing equities with EU domiciled equity fund portfolio holdings during 2021 – 2024 (as collated by Morningstar), the Report found that more than 50% of market value is concentrated in six of the Magnificent Seven tech companies with the balance made up of smaller companies, concluding that EU investment is directed at and spread among a diverse number of AI industry stakeholders. 

The Report noted that while there was an increase in portfolio shares in AI companies from 9% to 12% for passive funds and from 9% to 14% for active funds since 2023, the market value of these funds doubled in that time. The Report also noted that while more than a quarter of funds had no exposure to AI companies as of mid-2024, over the period 2021 to 2024, a number of funds increased their exposure to AI stocks in general and to the Magnificent Seven in particular. This contrasts with a more reactionary approach in the US in that period where most mutual funds dumped shares in AI stocks to avoid breaching regulatory or self-imposed diversification rules.

The Report concludes that an increasing investment in AI firms signals growing confidence in the sector, but that diversification is key to manage risks such as operational setbacks, regulatory challenges and changes in expectations. 

4. Keeping Under Review

The findings of the Report intimate that EU regulators will need to continually review developments in AI investment and use for systemic risk for the EU financial ecosystem. As noted at the outset, the expectation among industry stakeholders in Ireland is that EU regulators will continue with their proactive approach in developing and adapting legislation to meet market developments by way of the continual review provisioned in the EU AI Act with annual reviews by the European Commission to take place of types of high-risk AI systems and the prohibited AI practices, and a wider review to be taken every four years on amendments needed to extend the scope of the act and enhancements of the supervision and governance system. 

The Central Bank of Ireland has also been proactive in monitoring developments in AI and related crypto assets, with public addresses3 noting work to develop supervisory expectations of regulated entities using AI in financial services and confirming commitment to enforcement of the EU AI Act, with related work in the ongoing review of the Consumer Protection Code for digital financial services. The Central Bank has included in its Regulatory and Supervisory Outlook Report 2024 references to the use of AI in data processing and modelling and the continual risk posed by AI technology outpacing regulation and consumer protections put in place by regulators, which is to be continually monitored and addressed. Of particular interest is the Central Bank’s recently announcement of funding for PhD programmes in AI and data science with the aim of achieving safer financial systems and better outcomes for consumers of financial services. 

Talk To Us

If you would like to discuss or review any aspect of AI use in your enterprise, we would be delighted to connect with you.

For further information, please contact David Naughton (Partner and Head of Investment Funds and Financial Services Regulation), Narita Woods (Senior Associate), or another member of our Investment Funds and Financial Services Regulation Team.

Our Investment Funds and Financial Services Regulation hosted an event on The Adoption of AI in Financial Services in September 2024, with a keynote address by the then Minister of State for Financial Services Neale Richmond, T.D. 

Our advisory practice includes FinTech, crypto assets and services and AI in a variety of financial entity types regulated in Ireland.

1 International Monetary Fund (2024) Global Financial Stability Report – Steadying the course: Uncertainty, artificial intelligence, and financial stability ‘Advances in artificial intelligence: Implications for capital market activities.’