EMEAP's scam task force moves fraud data sharing from goodwill to supervision

A task force among central banks is a plumbing decision before it is a policy one. Central banks hold settlement data and supervisory returns. The records that actually identify a scam, such as mule account openings and the inbound transfer patterns around them, sit with commercial banks and payment operators. A channel between the 11 EMEAP members ^[7] can only carry what each member can compel its supervised institutions to hand over. That is the part that eventually reaches banks, and it reaches them as a reporting template rather than a communique.

The sequencing across the two days is the more useful tell. The scam information-sharing task force and the AI workstream were approved on July 23, alongside a training programme for junior central bankers ^[3][4]. The following day, AI adoption by financial institutions was discussed by governors together with heads of supervisory authorities, who exchanged what BSP described as best practices on responsible use ^[5][6]. One day apart ^[11]. Research departments circulate best practices; supervisory heads convert them into questionnaires and then into examination modules. Putting AI in that room is what separates this from a conference panel.

The thinness is worth stating plainly. Everything here is attributed to the Bangko Sentral ng Pilipinas, one of the forum's 11 member institutions ^[9], and the account carries no start date, no membership and no indication of which data types the scam channel will move ^[10]. Governors used the same meeting to discuss the macro-financial implications of regional supply disruptions ^[8], which is a reasonable indication of how much agenda weight any single item carried.

What survives that discount is still consequential for anyone running compliance across several of these jurisdictions. Crowdfund Insider's read is that the discussions point to rising regulatory expectations on AI governance, consumer protection and risk management as adoption accelerates ^[12], and the same report suggests cooperation may produce more consistent regulatory approaches without constraining adoption ^[14]. Consistency cuts both ways for a regulated firm. It lowers the cost of one AI governance framework across markets, and it removes the option of telling each supervisor a locally convenient story about model oversight.

The BSP account describes banks already using AI in customer service, fraud detection, risk management and operational processes ^[13]. Fraud detection is the overlap point: the same models that supervisors will want documented are the ones that would consume whatever the new channel produces. A bank that treats the AI inventory and the fraud reporting build as separate projects will pay for the integration twice. The cheaper working assumption is that a neighbouring supervisor's fields turn up in your own examiner's request, and that the first version of both arrives before the task force publishes anything with a name on it.