News about fraud

Mastercard is testing an AI-agent initiation probability score in the United States, part of its Agent Pay Trust Framework, designed to answer whether a transaction is being initiated by an AI agent at all, rather than only verifying who the agent is. The score draws on Mastercard Decision Intelligence, which processes roughly 143 billion transactions a year, and combines Skyfire's Know Your Agent identity technology with intent, behavioral and fraud signals, while Cloudflare provides web visibility into AI agent activity. The article cites a vendor-reported PYMNTS finding that 93% of merchants think AI providers should bear losses on failed transactions, while only 28% are willing to expose their full inventory to agents. It concludes that liability for erroneous agent purchases remains unresolved.