FIS–Anthropic Claude Opus 4.5 Partnership: AI for Financial Infrastructure Security

What happened

FIS, the New York Stock Exchange-listed financial technology group, has extended its partnership with AI safety company Anthropic, deploying Anthropic's latest model, Claude Opus 4.5 (referred to internally as Mythos 5), through a controlled-access programme called Project Glasswing. The initiative is specifically designed to apply frontier AI to the protection of software underpinning critical infrastructure.

Project Glasswing represents Anthropic's selective, invitation-based effort to direct its most capable models toward hardening the systems that societies depend on most — financial networks prominent among them. FIS, whose technology processes trillions of dollars in transactions annually, is among the first institutions to operate within this framework, using the AI to identify and address vulnerabilities in its own platforms.

Why it matters

For customer experience and service-design practitioners, this partnership is a reminder that the invisible layer of infrastructure reliability is itself a CX variable. When a payment fails, a banking app goes dark, or a fraud alert misfires, the customer's trust erodes — regardless of how polished the front-end experience is. Deploying frontier AI to secure the pipes that carry financial transactions is, in effect, an investment in the baseline expectation every customer holds: that the service will simply work.

From a behavioural-economics perspective, this matters because of loss aversion. Customers weight a service outage or security breach far more heavily than they credit equivalent periods of flawless performance. Institutions that treat infrastructure security as a CX discipline — not merely an IT obligation — are managing the asymmetric emotional calculus their customers apply every time they tap to pay or check a balance.

The Renascence take

Most commentary on AI in financial services fixates on the customer-facing layer: chatbots, personalisation engines, smart onboarding. The FIS–Anthropic extension quietly points to something more consequential — the use of advanced AI to protect the operational substrate that makes any customer promise credible in the first place.

The instinct in CX is to reach for AI as a tool for delight — faster responses, smarter recommendations, smoother journeys. But the deeper behavioural truth is that customers do not reward delight nearly as much as they punish failure. FIS is applying AI where the stakes are highest: not to impress customers, but to ensure the system never lets them down. Customer-obsessed operators should audit their own stack and ask an uncomfortable question — how much of their CX investment is decorating the front door while the foundations go uninspected? Security and reliability are not IT problems; they are the most fundamental service-design decisions an organisation makes.