Lloyds Banking Group is leaning into 21st century tech – yet trying to do so in a way that the data of its 28 million customers is kept away from untested AI models developers might be tempted to deploy.
Michelle Conway, lead data and AI scientist at LBG, last week told an audience of data scientists, DevOps pros and AI folk at an evening event by MLOps Community and JFrog, that “we’re very careful with your data – our security is huge. It’s like Fort Knox locked down.”
LBG – which looks after the cash of customers across Lloyds, Royal Bank of Scotland (RBS) and Halifax – traces its roots back to 1765, yet with many of its rivals jumping board the AI train, the fear of missing out on the latest tech craze is real.
In a positioning statement put out earlier this year, the bank said: “The integration of AI into financial services is not just about adopting new technology; it’s about reimagining the entire banking experience.”
Alas, as we’ve come to know, technology evangelists tend to have a “move fast and break things” attitude where security is somewhat lax – particularly when it comes to rushing out AI-powered tools.
“We’re very strict with our data and getting permissions,” said Conway. “I think that’s good because if you bank with us as a customer, you know your data is very secure, it’s not going to be exposed to a third party.”
This extends to banning devs from using AI model hosting frameworks like Hugging Face, because of the risk of staffers inadvertently downloading malicious or otherwise nefarious models placed in repositories by cyber ne’er-do-wells looking to steal sensitive data.
“Hugging Face is actually blocked in the bank, because we recognise that there’s cybersecurity attacks targeting that and we don’t want to risk it, so it’s been locked,” Conway explained. But she added that this isn’t necessarily a permaban.
“We’re looking at getting it unblocked soon, but it’s been a long project to get it through security and trying to get it through while keeping everything safe.”
Despite the caution with respect to some platforms, Conway said LBG is already implementing over 100 use-cases of AI across the group.
These include chatbots to provide assistance to both customers and staff, using AI for document processing, as well as using it to explore what the bank describes as improvements in digital transformation.
Conway kept her cards close to her chest on the exact details around these schemes but she detailed how Lloyds has deployed both Microsoft Co-Pilot and Google Gemini AI platforms.
“We’re very big on using Microsoft Co-Pilot, I could absolutely use it all day long, it’s amazing,” she said, and described Google’s service as “crucial” in aiding the bank to do behind-the-scenes work.
Earlier this month, Lloyd’s put out a paper examining AI in the financial sector, which claimed that banks are welcoming AI with open arms, with 60 percent of institutions reporting improved productivity thanks to initiatives around AI.
That, however, is in contrast to a UK government department’s recent three-month trial of Microsoft Co-Pilot, which concluded that the AI tool didn’t provide a clear productivity boost – sometimes performing tasks worse than a human would.
Microsoft recently admitted it is still trying to build return on investment cases for Copilot, however, with Barlcays getting 100,000 licenses earlier this year and LBG exploring its use in Power Apps and Power Automate, it seems the software giant may start to make bank from AI afterall. ®