FIS joins Anthropic cyber security project with Mythos 5

FIS has joined Anthropic's Project Glasswing to use the Mythos 5 artificial intelligence model in its internal security work. The move brings a major financial technology provider into a controlled-access cybersecurity initiative focused on critical software infrastructure.

FIS is testing Mythos 5 to scan and evaluate its own systems as part of the programme. It runs systems used for payments processing, money movement and core banking for thousands of financial institutions, placing its internal software security near the centre of day-to-day financial operations.

The arrangement is separate from FIS's commercial AI relationship with Anthropic. Instead, it forms part of the company's broader security effort, which includes industry intelligence sharing and engagement with sector bodies.

Project Glasswing is Anthropic's initiative for organisations that build or maintain software underpinning critical infrastructure. Participants use Anthropic's frontier AI models for defensive security work, with the stated aim of identifying and addressing vulnerabilities in widely used systems.

Anthropic has described FIS as one of more than 40 additional organisations given access under the scheme beyond the initial launch group. Those launch partners include large technology and security companies, a major bank and the Linux Foundation.

Security focus

For FIS, the project reflects the scale of the systems it operates. Its platforms are involved in clearing payments and running core banking functions, and a breach or software weakness in those systems could have implications beyond its own operations.

Mythos 5 is Anthropic's most advanced unrestricted model in its Mythos line and is reserved for approved organisations through Project Glasswing. Anthropic says the model was built for advanced cybersecurity, healthcare, and biology research, and that access is tightly controlled due to its potential to uncover serious vulnerabilities in operating systems.

Anthropic has also said a related model, Claude Mythos Preview, has already identified thousands of zero-day vulnerabilities across critical infrastructure. Under Project Glasswing, participating organisations use these models in research and defensive settings rather than in open public deployment.

FIS said its participation adds another layer to its existing security programme. It also cited work with FS-ISAC, the Financial Services Sector Coordinating Council, regulatory collaboration and industry intelligence sharing as part of its security posture.

Separate track

The distinction between this effort and FIS's commercial use of Anthropic tools is notable. Financial institutions and their technology suppliers have been expanding the use of AI in customer service, software development and internal operations, but security applications have drawn particular attention because they can help detect flaws while also raising questions about access, oversight and model control.

By keeping the Glasswing work separate from commercial deployment, FIS is signalling that this use case sits in a more restricted and specialised environment. Anthropic has framed Project Glasswing in similar terms, limiting access to organisations responsible for foundational software and to defensive security work.

The wider backdrop is a race across the technology and financial sectors to assess how advanced AI models can be used in cyber defence. Companies are seeking ways to automate parts of vulnerability discovery and code review while avoiding new operational or governance risks.

Anthropic has committed up to USD $100 million in usage credits and USD $4 million in donations to open-source security organisations to support the initiative. It has been said that lessons from the project will be shared more broadly with the industry.

That may be of particular interest in financial services, where cyber resilience requirements have tightened and where software supply chains, third-party dependencies and interconnected payment systems can quickly turn a technical weakness into a broader operational issue.

FIS did not disclose the scale of its testing or any specific findings from its use of Mythos 5. It said only that it is actively using the model through Project Glasswing to assess its own systems.

The company said the initiative reflects "the same disciplined approach to applying advanced AI in financial systems where security, reliability and trust are essential."