Lyrie Raises $2M Pre-Seed and Open-Sources Agent Trust Protocol to Secure the Autonomous AI Economy
Dubai-based Lyrie.ai emerged from stealth with a $2 million pre-seed round and the release of Agent Trust Protocol, an open cryptographic standard for identifying, scoping and revoking autonomous AI agents that the company plans to submit to the IETF.
Lyrie.ai, an autonomous-cybersecurity startup spun out of Dubai-based OTT Cybersecurity LLC, exited stealth on Monday with a $2 million pre-seed round and an unusually ambitious opening play: a royalty-free, open cryptographic standard for AI agents that the company says it will submit to the Internet Engineering Task Force. Founder and chief executive Guy Sheetrit, a veteran offensive-security researcher, said the timing is deliberate. Agentic AI products are being shipped into production at enterprise scale, but the identity and trust plumbing those agents need is still being assembled in real time.
The protocol, dubbed Agent Trust Protocol (ATP), defines a common scheme for agent identity, scope enforcement, cryptographic attestation, delegation of authority between agents and revocation of compromised credentials. A reference implementation is already live on GitHub under an MIT license, and Lyrie is positioning ATP as deliberately vendor-neutral, hoping to attract competing model providers and orchestration frameworks before any one of them ships a proprietary alternative. The IETF submission is intended to anchor the standard outside of any single company's control.
Alongside the protocol, Lyrie is bringing a commercial platform to market that pairs autonomous penetration testing with adversarial AI red-teaming. The system can run a seven-phase pen test from a single command on infrastructure ranging from consumer hardware to H200 GPU clusters, and it ships with coverage tuned to the OWASP ASI 2026 threat catalog, including zero-day research in compiled software. OTT Cybersecurity has also been admitted to Anthropic's Cyber Verification Program, granting Lyrie sanctioned access to run vulnerability research and red-team workflows on Claude infrastructure within the lab's safety guidelines.
The pre-seed capital will go toward expanding Lyrie's security-research team, scaling the IETF submission process and seeding deployments with early enterprise and government customers. The company says it is already preparing a Series A aimed at financing a wider commercial rollout. With Sierra, Cognition and a handful of other agent-platform vendors now selling autonomous workers to Fortune 500 buyers, Lyrie is betting that the next billion-dollar layer of the AI stack will not be a smarter model but the cryptographic scaffolding that decides what those models are allowed to do.
