Bank of England: We can’t eliminate bias in AI
A top boss at the Bank of England has warned the use of artificial intelligence programs will never be bias-free and instead users must be aware of the risk.
Jem Davis, chief compliance officer at the Bank of England, said: “My view is you’re probably not going to be able to eliminate bias completely, but I think it is something you can actively manage.”
Speaking at the City and Financial Global’s Data, AI and the future of Financial Services summit, Davis said having numerous units across a whole organisation looking into data can help “spot patterns where bias can be more visible”.
The comments came as Davis was quizzed on how banks can ensure their data is “AI-ready” without baking in the prejudices of the past.
She added: “If you’ve got a good governance process, then you’re going to have programs in place for testing models with diverse data sets, you are monitoring outcomes…
“That’s why I would say on it… the objective isn’t [bias] free. It’s sort of being bias aware.”
Regulators face AI scrutiny
The Bank of England – along with other regulators – has come under fire for leaving the financial system exposed to “serious harm” from AI risks.
A group of influential MPs on the Treasury Select Committee took aim at the institutions for “not doing enough to manage the risks presented by AI,” despite claims from regulators that the current rulebook was sufficient.
“I do not feel confident that our financial system is prepared for a major AI-related incident,” Chair of the Treasury Select Committee Dame Meg Hillier wrote in the conclusion.
“This needs to be addressed.”
Recent regulatory progress on AI has seen the UK’s financial watchdog launch the Mills Review, a major probe into “agentic AI” and the long-term systemic risks of autonomous systems acting in retail markets.
The Financial Conduct Authority is also scouting for its next round of AI Live Testing. It follows the regulator opening its doors in December for financial firms to trial AI in a controlled environment in December without the fear of tripping regulatory wires.
NatWest, Monzo, Santander, Scottish Widows, Gain Credit, Homeprotect and fintech Snorkl were among those participating in the event, which the regulator said would help fix governannce before the algorithms begin making high-stakes decisions.
Speaking at the London conference on Monday, Davis said: “It’s going to be key to design structures that allow safe experimentation without compromising control and the science background.”
She added: “There’s a big new thing and it is moving really fast, but there are structures that we can put in to face into those headwinds quite strongly.”
The Bank’s Financial Policy Committee has previously warned a crash in the roaring value of US tech giants could spark trouble overseas after comparing the surge in AI-related stocks to the peak of the dot com bubble.