Codenotary Joins Anthropic Verified Access Program to Advance Autonomous AI Security Research

Participation strengthens Codenotary’s ability to research emerging AI agent threats and accelerate protections for enterprises

HOUSTON--(BUSINESS WIRE)--Codenotary, leaders in assuring safe and secure use of AI, today announced it has joined Anthropic’s Cyber Verification Program, a security research initiative that enables qualified organizations to conduct advanced defensive AI cybersecurity research under Anthropic’s responsible-use framework.

As enterprise adoption of AI agents accelerates, organizations are moving beyond isolated chatbots to deploy thousands of autonomous systems with persistent memory, delegated authority, and direct access to internal applications, APIs, cloud infrastructure, source code repositories, and sensitive business data. This shift is fundamentally expanding the enterprise attack surface.

Traditional cybersecurity models were built to defend deterministic software. They were not designed for AI systems capable of independent reasoning, tool invocation, agent-to-agent delegation, or dynamic workflow adaptation.

“AI has fundamentally changed how attacks are created, refined, and executed,” said Moshe Bar, CEO and co-founder, Codenotary. “To secure autonomous AI, we need to understand how these systems can be manipulated before attackers operationalize them. The future of cybersecurity will be defined by speed, reducing the time from months to hours between innovation, attack discovery, and protection.”

Through Anthropic’s Cyber Verification Program, Codenotary can accelerate defensive research into emerging AI-native attack vectors, including multi-agent compromise, prompt-chain manipulation, memory poisoning, indirect prompt injection, Model Context Protocol (MCP) abuse, tool hijacking, delegated trust failures, identity spoofing, and autonomous lateral movement.

Unlike traditional threat research, which often analyzes incidents after deployment, Codenotary’s approach focuses on proactive adversarial research designed to identify future attack patterns before they reach production.

This research directly strengthens AgentMon, Codenotary’s adaptive AI runtime security platform, which currently observes and protects more than 5 million AI agent interactions daily across enterprise environments.

AgentMon continuously builds behavioral baselines for AI agents, dynamically generates runtime security policies, detects anomalous agent behavior, and identifies emerging attack patterns in real time. This enables organizations to detect not only known threats, but entirely novel forms of agent misuse.

That distinction is increasingly critical as AI systems gain the ability to create new attack paths using legitimate enterprise tools, without relying on malware or previously documented exploits.

About Codenotary

Used by hundreds of customers worldwide – including the world’s leading banks, governments, and defense organizations – Codenotary delivers technology that enables secure and trusted agentic networks in the modern organization. For more information, visit www.codenotary.com.