AI is becoming a budget and security-control problem for payments firms, with Ramp and FIS moving in different directions to manage the same operational pressure, Banking Dive reported on July 22, 2026.

Ramp introduced an AI spend management tool for finance teams, while FIS joined Anthropic's Project Glasswing to test a model against payment-infrastructure vulnerabilities.

AI spending can now sit across departments, employees, projects and external model providers rather than one software contract.

Security teams face a different version of the same problem: model use adds new code, workflow and infrastructure surfaces that payment processors must monitor before failures reach customers.

Ramp Adds AI Spend Management For Finance Teams

Ramp said in its July 16 release that the new tool tracks spending with providers including OpenAI, Anthropic and Gemini.

It aggregates AI expenditure by company, department, employee or project, giving finance teams a view of token usage before costs are consolidated into invoices.

The workflow is built around limits and alerts.

Finance teams can restrict token usage by category, receive notices when spending surges and warn employees before overruns occur.

Ramp's release placed the product inside a fast-growing cost category, saying AI token spending had become the fastest growing area of business spend while finance teams had little visibility into how the money was being used.

The controls create a budget-governance layer for model use, while customer-level savings, deployment counts and evidence that the alerts changed employee behaviour remain outside the public materials.

FIS Tests Anthropic Model For Payment Infrastructure Security

The Anthropic partnership is aimed at security, not finance visibility.

FIS's July 17 release named Project Glasswing, Anthropic's initiative for finding software vulnerabilities, and placed the payments processor inside a programme that Anthropic's website also lists with Google, JPMorgan Chase and Amazon Web Services.

Mythos 5 is the model FIS named for the test.

The payments processor is using it to protect payments infrastructure, a different operating surface from Ramp's spend dashboard because payment processors handle transaction systems where software weaknesses can become fraud, outage or compliance risks.

FIS already has a separate commercial relationship with Anthropic for AI agents.

The Project Glasswing work adds a security-testing layer to that relationship, but the public material does not include vulnerability counts, benchmark methodology or post-testing remediation data.

AI Controls Move From Experimentation To Operating Risk

Together, the deployments put AI controls into two established finance workflows: managing technology expenditure and testing payment software for weaknesses.

Ramp is addressing consumption before costs reach the invoice stage, while FIS is applying a model to a payment-security process.

The company materials do not provide audited savings, customer adoption figures, vulnerability counts or measured security outcomes.

Those results will determine whether the tools change operating performance beyond their initial launches.