Lyrie.ai Enters Anthropic's Cyber Verification Program with New Agent Trust Protocol

OTT Cybersecurity LLC, the Dubai-based developer behind Lyrie.ai, has been accepted into Anthropic's Cyber Verification Program (CVP) as part of the initial cohort. The announcement, released May 11, 2026, coincides with the public launch of the Agent Trust Protocol (ATP), an open cryptographic standard designed to verify AI agent identity, scope, and authorization in real time.

The CVP acceptance validates Lyrie's work around vulnerability research, offensive security tooling, and red-team workflows on Claude's AI infrastructure. This verification pathway allows the company to operate within Anthropic's safety and security policies while conducting dual-use cybersecurity operations.

According to the official press release, CEO and Founder Guy Sheetrit positioned the company as foundational infrastructure for the agentic AI era. "Lyrie isn't a security tool that sits alongside AI. It's the security layer that AI runs on top of." The distinction matters—most security vendors bolt on scanning after deployment. Lyrie is attempting to embed verification into the agent execution layer itself.

The Agent Trust Protocol covers five cryptographic primitives: identity (who the agent is), scope (what it can do), attestation (whether it's been tampered with), delegation (who authorized it), and revocation (whether authority remains valid). The protocol is open, royalty-free, and slated for submission to the Internet Engineering Task Force (IETF). A reference implementation is published under MIT license at github.com/OTT-Cybersecurity-LLC/lyrie-ai.

Independent coverage from Cybersecurity Insiders corroborates the timeline and technical scope. The article notes that enterprises and governments are deploying autonomous AI agents at unprecedented speed—agents that read mail, write code, move money, and sign contracts. The security model for those agents has not existed at enterprise scale until now.

Every AI agent on the internet today is effectively anonymous. You don't know who it is, what it's authorized to do, or whether it's been tampered with. ATP is designed to change that dynamic by providing real-time cryptographic verification. Think of it like SSL/TLS for the web, but for AI agent communications (which is exactly how Sheetrit frames it, though the analogy is apt).

The Lyrie platform itself delivers autonomous offensive and defensive security across the full threat lifecycle. Capabilities include "lyrie hack"—a single command that runs a seven-phase autonomous penetration test, generating proof-of-concept exploits and code-level remediation. The platform also features GPU-accelerated red-teaming using GCG and AutoDAN adversarial workflows on H200 GPU infrastructure, OWASP ASI 2026 coverage, and Omega-Suite binary research for autonomous zero-day discovery.

Hardware-flexible deployment ranges from consumer hardware through enterprise GPU clusters. Nine built-in security tools handle reconnaissance, exploitation, and remediation in a single agent. This consolidation reduces the friction of switching between multiple security platforms—a practical consideration for teams already drowning in tool sprawl.

Alongside the CVP acceptance, Lyrie completed a $2 million pre-seed funding round and officially exited stealth mode. The raise will fund expansion of the security research team, infrastructure scaling, the IETF submission process for ATP, and deeper enterprise and government partnerships. The company is preparing a Series A round to scale deployment across enterprise and government markets.

Anthropic's Cyber Verification Program itself signals a broader industry shift. The company has announced new cyber safeguards that automatically detect and block requests suggestive of prohibited or high-risk cybersecurity uses. The CVP is Anthropic's mechanism for pre-positioning governance around organizations whose work might trigger these safeguards. Translation: Anthropic knows frontier AI will be used for both attack and defense, and they're creating a verified pathway for legitimate operators.

The timing aligns with Anthropic's May 10, 2026 launch of Claude Security in public beta, embedding Claude Opus 4.7 directly into security platforms from CrowdStrike, Palo Alto Networks, Microsoft Security, SentinelOne, Wiz, and TrendAI. The compression of the vulnerability discovery-to-exploitation timeline is now measurable in seconds, not days. CISOs no longer have a choice: autonomous vulnerability scanning is table stakes.

Lyrie is also exploring similar verification pathways with other leading AI labs as part of its mission to build trusted security infrastructure for autonomous AI systems. This multi-lab approach suggests the company is positioning ATP as a universal standard rather than an Anthropic-specific solution.

Whether enterprises actually adopt ATP at scale remains the real question. Open standards succeed only when they solve a problem better than proprietary alternatives. The cryptographic primitives are sound, but adoption depends on whether security teams will integrate ATP into their existing agent frameworks or treat it as another checkbox requirement.

The $2 million pre-seed round buys Lyrie runway to prove the concept. But in the agentic AI economy, infrastructure that doesn't become infrastructure is just another tool that gathers dust. Time will tell if ATP becomes the trust layer underneath the agentic AI economy—or if it becomes another well-intentioned protocol that nobody actually implements.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn