FCA urged to regulate AI pension advice as algorithmic risks mount

The Financial Conduct Authority is facing pressure to establish regulatory frameworks for artificial intelligence in pension advice, as industry figures warn that inadequately supervised algorithms could lead to poor retirement outcomes for consumers.

Broadstone, a financial services consultancy, has called on the UK regulator to develop specific guidance for AI deployment in pension decision-making, according to Financial Planning Today. The intervention comes as financial institutions increasingly adopt machine learning systems to automate retirement planning recommendations, raising questions about accountability when algorithms produce suboptimal advice.

The regulatory gap represents a significant compliance challenge for enterprises operating in financial services. Whilst the FCA has issued general guidance on algorithmic accountability, pension-specific AI applications remain largely unaddressed in current frameworks. This creates uncertainty for firms investing in automation whilst simultaneously exposing consumers to potential harm from biased or poorly trained models.

The timing of Broadstone’s intervention reflects broader regulatory momentum across financial services. The European Union’s AI Act, which came into force in August 2024, classifies AI systems used in credit scoring and insurance underwriting as high-risk applications requiring stringent oversight. UK regulators have taken a more principles-based approach, but pressure is mounting for sector-specific rules as AI adoption accelerates.

Pension decisions present particular algorithmic challenges. Unlike simpler financial products, retirement planning requires models to account for decades-long time horizons, complex tax implications, and highly individualised risk tolerances. An AI system trained predominantly on historical market data may fail to account for structural economic shifts or produce recommendations that systematically disadvantage certain demographic groups.

The business implications cut across multiple stakeholders. Traditional financial advisers could benefit from clearer regulatory boundaries that emphasise the continued need for human oversight, potentially slowing the shift towards fully automated advice platforms. Conversely, fintech firms building AI-driven pension products face increased compliance costs and longer development cycles if prescriptive rules emerge.

For pension providers and workplace schemes, the regulatory uncertainty complicates technology procurement decisions. Firms must balance the efficiency gains from automation against the reputational and financial risks of deploying systems that later prove non-compliant with evolving standards. Insurance against algorithmic failures remains an immature market, leaving enterprises exposed.

The FCA’s response will likely influence regulatory approaches beyond the UK. Financial services regulators globally are grappling with similar questions about AI governance, and Britain’s historically influential position in financial regulation means its framework could serve as a template for other jurisdictions.

Industry observers anticipate the FCA will favour a principles-based approach rather than prescriptive technical standards, consistent with its broader regulatory philosophy. This would likely focus on outcomes—requiring firms to demonstrate that AI systems produce fair results and that consumers understand when algorithms influence their advice—rather than mandating specific technical architectures.

The regulatory examination also intersects with the government’s broader pension policy objectives. The Department for Work and Pensions has been pushing for increased pension participation and better retirement outcomes, particularly among younger workers. AI tools could theoretically advance these goals by making personalised advice more accessible, but only if systems prove reliable and trustworthy.

What remains unclear is whether the FCA will establish a dedicated AI oversight function or integrate algorithmic supervision into existing regulatory structures. The authority’s resource constraints and the technical complexity of auditing machine learning systems suggest meaningful enforcement will require significant capability building.

Market participants should monitor the FCA’s upcoming policy statements on operational resilience and consumer duty, both of which could incorporate AI-specific expectations. Firms deploying pension algorithms would be prudent to document model development processes, establish human oversight mechanisms, and conduct regular bias audits ahead of formal regulatory requirements.

The Broadstone intervention signals that industry self-regulation may prove insufficient, making formal FCA guidance increasingly probable as AI adoption in retirement planning continues to expand.