By PAUL O’DONOGHUE, Senior Correspondent
THE best LLMs (large language models) get 1 in 3 financial crime risk ratings wrong, a new study has found.
Cove Labs, a firm which specialises in evaluating AI models, said it tested four leading frontier AI providers across 110 expert-authored fraud scenarios. It evaluated whether the LLMs could identify potential financial crime issues and recommend the right next steps.
Subscribe now to have unlimited access
With our membership subscription, you will have unlimited access to the AML Intelligence site, updated daily with the latest analysis, opinion, and breaking news across the sector, newsletter delivered twice per week, access to our Global Bank Fines & Penalties database, free access to Boardroom Series and more!
