The Monetary Authority of Singapore is validating artificial intelligence models for detecting financial fraud using real bank account information and transaction data. It is running a joint project with Singapore’s Government Technology Agency and the police to improve fraud detection. MAS is pooling historical transaction data from five banks and training and evaluating models using participating banks’ actual account numbers. It also set up security measures, including hashed account numbers and controlled data-sharing environments.

As AI coding tools rapidly improve, more developers are pushing code into production without reviewing it line by line, widening debate over how far responsibility for AI-generated code should extend. Developer Simon Willison says the line between “vibe coding” and agent-based engineering is blurring as tools like Claude handle tasks from SQL queries to tests and documentation. He warns that rising trust can weaken review and accountability, and shift how code quality is judged.