TL;DR

The Regulator of Social Housing’s deputy chief executive, Kate Dodsworth, has set out how the regulator will treat landlords’ growing use of AI in a comment piece published by Inside Housing on Tuesday 13 October. The regulator will not write AI rules of its own, she says, because its framework judges outcomes, but it “will intervene only where the outcomes required by our standards are compromised”. It joins a run of UK regulators stating positions on AI in recent days, including the senior judiciary and the Solicitors Regulation Authority.

An outcomes test, not a rulebook

Dodsworth describes AI use across social housing as expanding quickly, from customer service and repairs to data analysis and communications, and cites recent sector research on the pace of take-up. The upside, in her account, is landlords working faster, using data better and identifying risk earlier in areas such as asset management and tenant insight.

The regulator’s response is to apply its existing standards rather than add new ones. Those standards already cover complaints and redress, fairness, governance, risk management and transparency, all of which she says bear on AI. The one principle she singles out is that human judgement stays essential: a model can analyse and advise, but decisions affecting people’s homes still need an accountable person behind them.

The risks she names

The piece asks landlords to have a robust framework and proper oversight in place before any significant adoption, and to manage cyber security, bias, data protection and data quality as they go. Badly designed tools could give tenants a worse experience, she warns, and inaccurate data, model errors or algorithmic bias could produce unfair outcomes. Digital exclusion and accessibility could shut some tenants out of AI-enabled services altogether, and she argues trust depends on landlords being open about when AI is in play and whether it is supporting a human decision or standing in for one.

Looking forward

RSH will monitor adoption across the sector and says its own approach and staff capability will develop as use spreads. Responsibility, though, sits with the landlord. The stance mirrors two recent statements: senior judges said AI may assist but must never decide a case, and the SRA opened a consultation on writing AI into solicitors’ competence standards. For housing associations and the suppliers selling them repairs triage or tenant chatbots, the practical reading is that an AI failure will be assessed as a standards failure, with no separate AI defence. In our view, the line to watch is the demand for a framework before significant adoption, which puts governance ahead of the pilots many landlords are already running.