Lyrie raises $2M to secure the AI agent era
What's the deal? Lyrie.ai, the autonomous cybersecurity platform built by OTT Cybersecurity LLC, has closed a $2M pre-seed round and exited stealth. The Dubai-based company is building what it calls the security layer for autonomous AI agents — software that acts on behalf of humans across sensitive systems.
The funds will go toward expanding Lyrie's security research team, scaling infrastructure, and pushing its Agent Trust Protocol (ATP) through the Internet Engineering Task Force (IETF) standardisation process. The company is already preparing a Series A to scale into enterprise and government markets.
Alongside the funding, Lyrie released ATP as an open, royalty-free cryptographic standard for AI agent identity, scope enforcement, attestation, delegation, and revocation. The reference implementation is available under an MIT licence.
Why now? Enterprises and governments are deploying autonomous AI agents faster than security frameworks can keep up. These agents read email, write code, execute transactions, and operate across sensitive systems — often with broad access and minimal oversight.
"Every AI agent on the internet today is effectively anonymous. No identity verification, no scope enforcement, no tamper detection," said Guy Sheetrit, chief executive officer and founder of OTT Cybersecurity LLC. "We built the infrastructure that changes that — and we built it as an open standard so the entire industry can adopt it."
OTT Cybersecurity has also been accepted into Anthropic's Cyber Verification Programme, a framework for verified dual-use cybersecurity operators. That gives Lyrie access to red-team workflows on Anthropic's Claude infrastructure.
What could go wrong? Building an open standard is one thing; getting the industry to adopt it is another. ATP aspires to become the trust layer for AI agents the way SSL/TLS became the trust layer for the web — a bold analogy for a pre-seed startup. Competing standards could emerge from larger players, and IETF submissions can take years to finalise.
The platform's offensive capabilities — autonomous penetration testing, adversarial red-teaming, zero-day research — also raise dual-use questions, even with Anthropic's safety guardrails in place.
The signal: As AI agents gain autonomy, the market for securing them is becoming a category of its own. Lyrie's bet is that agent identity and trust will need dedicated infrastructure, not just bolted-on features from existing security vendors. If agentic AI scales as fast as the industry expects, the companies that define its security primitives early could shape the entire ecosystem.
Read more: globenewswire.com
