Wallarm announced its role in the publication of “A2AS: Agentic AI Runtime Security and Self-Defense,” a groundbreaking research project led by Eugene Neelou (OWASP, Wallarm) together with researchers from AWS, Bytedance, Cisco, Elastic, Google, JPMorganChase, Meta, and Salesforce.

The A2AS framework introduces a new security layer for AI agents, LLM-powered applications, and AI protocols, similar to how HTTPS secures HTTP.

Cyber Technology Insights : Kyivstar Launches Locally Hosted AI Platform In Its Cloud

The A2AS framework is built on three breakthrough capabilities that fundamentally address agentic AI security risks such as prompt injection, tool misuse, and agent compromise:

  • Behavior Certificates: The industry’s first mechanism for declaring and enforcing AI agent actions and permissions. Like HTTPS certificates secured the web, behavior certificates can secure agentic AI interactions with users, tools, and other agents.
  • Model Self-Defense Reasoning: Embeds security awareness directly into the AI model’s context window, guiding it to recognize and reject malicious or untrusted instructions in real time without any external components or guardrails.
  • Prompt-Level Security Controls: Provides authenticated prompts, security boundaries, and policy-as-code so that every request and interaction is verified, sandboxed, and aligned with enterprise security policies.

As enterprises rapidly deploy agentic AI into workflows across finance, healthcare, and infrastructure, the security risks scale from individual task failures to enterprise-wide compromise. Traditional guardrails and post-processing methods have proven to be too slow, too complex, and too costly. A2AS offers a practical, lightweight, and scalable approach that protects AI agents at runtime without adding latency or operational complexity.

Cyber Technology Insights : Murex Announces Strategic, Multi-Year, Global AWS Collaboration Agreement

Eugene Neelou, an industry pioneer and Head of AI Security at Wallarm, serves as the lead for the A2AS project. Neelou previously coined the term MLSecOps, co-founded the world’s first AI red teaming startup, and co-authored the OWASP Top 10 for LLM Security. He is joined by Ivan Novikov, Founder and CEO of Wallarm, who contributed his expertise in API and AI security.

“AI agents are already in production, and they introduce a dangerous new attack surface,” said Ivan Novikov. “With A2AS, we’ve shown that security can be embedded directly into the agent runtime, turning self-defense from a theory into a practical defense layer.”

“AI agents are rapidly infiltrating enterprise software, requiring privileged access and deep integration with company-wide tools,” added Eugene Neelou. “Under pressure to adopt AI, enterprises are voluntarily incorporating vulnerable-by-design AI capabilities. Without deliberate security hardening, disaster is inevitable.”

Cyber Technology Insights : ThreatBook Launches Best-of-Breed Advanced Threat Intelligence Solution 

Source: prnewswire

To participate in our interviews, please write to our CyberTech Media Room at sudipto@intentamplify.com