Artificial Intelligence – Humanity's sunset or the dawn of a golden age for financial services?

Listen now or read the full briefing below

KEY INSIGHTS

Regulators favour evolution over revolution: The FCA's principles-based approach means that UK private capital firms face no new AI rulebook, but must apply existing governance frameworks more rigorously as AI tools become increasingly autonomous.

Governance and resilience demand urgent attention: As AI deployment deepens, firms must map their exposure to AI supplier concentration, invest in cybersecurity, and ensure that senior managers — not machines — remain clearly accountable for decisions.

It's a case of when, not if: Despite alarming predictions from within the AI industry about the risks, the productivity gains from disciplined AI deployment are real, and firms that build the right governance and skills infrastructure now will hold a lasting advantage.

A regular briefing for the alternative asset management industry

Ten percent. That is the probability one safety researcher at Anthropic, one of the world's largest artificial intelligence companies, recently assigned to the annihilation of humanity due to AI. Such predictions make headlines. Cynics might suggest that they redirect focus away from the real, if somewhat more prosaic, practical and ethical issues arising from the everyday adoption of AI. If people are talking about human extinction, they are not discussing concerns such as invisible algorithm-driven discrimination, intellectual property infringement, or AI's implications for national security.

Nonetheless, noisy messages about the risks of AI usage should not distract private capital firms from a more pressing question: how to deploy AI responsibly, and how to reap its considerable rewards.

The UK financial regulator's answer is encouraging. The Mills Review, published in July and commissioned by the UK Financial Conduct Authority, examined how AI could transform financial services by 2030. Its focus was retail finance, but its conclusions apply more broadly. The FCA's existing rules, it found, remain fit for purpose. There is no dedicated AI rulebook on the horizon. The FCA continues to rely on its established, principles-based framework — broad rules on governance, accountability and operational resilience — rather than prescribing detailed conduct requirements for every new technology. For an industry already managing a demanding regulatory workload, that restraint is welcome.

This is good news for GPs. The EU has taken the opposite approach with its AI Act, imposing detailed and prescriptive requirements on firms operating in Europe. The FCA has chosen supervised evolution instead. It expects firms to apply existing governance frameworks to AI risk, not to wait for new ones. That is a sensible position. Over-regulation at this stage risks stifling innovation and reducing competition and economic growth. The UK should hold its nerve.

That said, the Review makes clear that the FCA's light touch approach has limits. Where AI tools operate with limited autonomy — where a human remains in control as operator, collaborator or decision-maker — existing governance rules are broadly adequate. This describes the operational set-up of most private capital firms today. But as AI becomes more autonomous, the governance challenge sharpens. Where humans become mere approvers, rubber-stamping decisions made by machines, demonstrating accountability becomes difficult. "The AI did it" will not satisfy a regulator. Senior managers remain responsible.

As we set out in an earlier issue of Alternative Insights, the FCA has been signalling this expectation for some time and in many ways, it is not new. Firms deploying algorithmic trading techniques have been expected to understand the activities of their algorithms and to ensure appropriate human oversight for many years. However, the scope and capabilities of general purpose AI tools powered by large language models are several orders of magnitude larger than a trading algorithm. Concerns that were once the preserve of a handful of quants are now becoming part of everybody's job description.

"The question for private capital firms is not whether to deploy AI. It is when and how to do it in a way that results in effective management of the risks and clear accountability."

Looking further ahead, the emergence of agentic AI — tools empowered to take sequences of autonomous actions within defined parameters — will require a more fundamental mindset shift. As AI agents become more capable, the role of human staff will need to be clearly defined. When should a human step in? How should they document their challenge to an AI's actions? Where do they escalate concerns? These are not abstract questions. They require firms to think carefully about skills, training and organisational design. The labour market may not yet have the talent to fill every role this new world demands, and firms that start building those capabilities now will have a meaningful advantage over those that do not.

Alongside governance, operational resilience is a key concern. The risks begin the moment a firm deploys any AI tool, and they grow as usage deepens. A joint FCA and Bank of England survey in 2024 found that the top three AI providers dominate the supply of cloud services, models and data to the UK financial sector. That concentration creates systemic risk. If a major AI supplier fails, firms that rely on different products from the same underlying provider may find that their backup solutions offer no real protection. GPs need to map their AI exposure carefully — both direct and through third-party service providers — and stress-test their contingency plans accordingly.

Cybersecurity is a related threat. A joint statement from UK regulators published in May 2026 warned that AI is making fraud and cyber-attacks faster, cheaper and more scalable. The Mills Review reinforces this. AI models can identify and exploit weaknesses in firms' IT systems with growing speed and sophistication. Investment in detection and response capabilities is not optional. Firms that fail to keep pace face mounting exposure, and the FCA has made clear that senior managers within firms are expected to identify and prioritise the remediation of the most pressing security risks.

Firms will also need to invest more in managing AI suppliers. Pricing, access to appropriate models and data sovereignty will all become central to commercial strategy. So will concentration risk. Geopolitical pressures are already shaping which AI products are available in which markets, and that dynamic is unlikely to ease. Firms that become overly dependent on a single provider expose themselves to both commercial and political risk. Supplier relationships that were once routine procurement decisions are becoming strategic ones.

The macroeconomic picture adds another layer of complexity. In April 2025, the UK's Financial Policy Committee warned that widespread AI adoption in investment decisions could drive herding behaviour and correlated trading, amplifying shocks in a downturn. It also raised the possibility of AI-enabled market manipulation potentially occurring undetected by humans, and becoming self-reinforcing as models learn. More broadly, AI may disrupt the economic viability of entire business sectors, unsettling conventional assumptions about credit and market risk across private portfolios. Fund managers will need to factor AI-driven disruption into their investment cases, not treat it as a background risk to be noted and set aside.

None of this supports the argument for widespread catastrophising about AI adoption. Instead, it reinforces the case for preparation. The firms that will benefit most from AI are those that build the governance, invest in the infrastructure and develop the human expertise to deploy it well. The potential productivity gains are genuine. The possibility of driving desperately needed UK economic growth is real. Autonomous AI, managed with discipline, could transform the economics of running a private capital firm.

The existential warnings from the AI industry deserve attention. A 10% probability of world-wide disaster, postulated by an industry insider, is clearly noteworthy. But at their heart, these predictions reinforce the arguments for controlled, thoughtful use of this rapidly evolving technology and careful consideration of its wider implications. The question for private capital firms is not whether to deploy AI. It is when and how to do it in a way that results in effective management of the risks and clear accountability. With the right answers, the industry will be able to bask in the white heat of the new digital age.

Read Samuel Brewer Profile