Visa will offer an inside look at Project Glasswing and how the most powerful agentic models are changing enterprise security at VB Transform 2026

Visa’s Tech Team Uncovers Security Risks in Advanced AI Models

Visa’s technology team recently delved into the world of advanced AI models with Anthropic’s Mythos model, and the security implications were immediately apparent. Rajat Taneja, Visa’s president of technology, shared insights from the team’s Project Glasswing, which highlighted the rapid identification and exploitation of vulnerabilities in critical code bases by attackers.

As one of the chosen companies to test Anthropic’s forthcoming model, Visa’s findings underscored the widening security gaps in enterprise systems and the potential risks posed by malicious actors. Taneja emphasized the importance of autonomous defenses in the age of AI, where attacks can now operate autonomously, creating a significant asymmetry in the security landscape.

According to Cisco’s State of AI Security 2026 report, threat actors now have access to powerful AI agents that can automate cyberattacks at a scale and speed beyond human capabilities. To address these evolving threats, Visa is implementing abstraction layers, observability measures, and data guardrails to secure its autonomous commerce frameworks. Additionally, the company has introduced an open-source, AI-driven security framework to streamline vulnerability discovery and remediation processes.

Visa’s proactive approach represents a necessary shift for enterprise IT teams to safeguard against evolving threats from autonomous agents controlled by bad actors. Taneja will delve into these insights and technical details during his session at VB Transform, titled “Inside Project Glasswing and Mythos: Securing the agentic future today,” scheduled for July 15.

The event will feature other sessions focused on agentic AI security, including discussions on open-source proxies for securing critical flaws, building trust layers for autonomous B2B commerce, and strategies for securing high-stakes transactional systems. Additionally, a panel discussion will explore best practices for permissioning, sandboxing, and incorporating human-in-the-loop controls in AI security.

Interested in attending VB Transform 2026? Register here. Complimentary passes are available for senior technology leaders. Contact the event organizers to secure your spot at this insightful gathering.