Everyone is talking about AI, but I think it’s important to ensure those who are talking understand it and how to use it. This was a strength of todays AI Innovation showcase because we heard from people that are looking at use of AI to tackle cyber security risks. A great range of speakers brought us up to date on current challenges, including:
Stevie Mullaghan of Microsoft,
Darnell Waker, Commercial Lead for Banking and Payments at Sumsub,
Thomas Knowles, Head of Security Operations at Clear.bank
Debs Goff, Fraud Risk Manager at HSBC
Mason Allen, Head of Growth & Partnerships at Reality Defender
My overall take away was to constantly challenge your control environment - to test security sounds obvious but how often do you do it? Sadly it needs constant attention.
AI LLMs do have weaknesses in the security and an area of well known vulnerability related to 'prompt injection', which is when a user creates an input designed to make the model behave in an unintended way. With the ever growing flow of data to third-party applications and services, the risks from malicious prompt injection are something to keep in mind.
