UAE-Based Cybersecurity Startup Lyrie.ai Exits Stealth with $2M Seed to Build Security Infrastructure for Autonomous AI Agents

The agentic AI economy has a trust problem: and a Dubai-based startup has stepped forward to solve it.

Lyrie.ai emerged from stealth on 14 May 2026 with a $2 million seed round, targeting one of the fastest-growing and least-addressed vulnerabilities in enterprise AI: the absence of a foundational identity and trust layer for autonomous AI agents.

As businesses across the GCC and beyond accelerate deployment of agentic AI; systems that don't merely generate text but take actions, access databases, execute transactions, and coordinate with other agents; the security architecture underpinning those agents has lagged dangerously behind. Lyrie.ai's founding thesis is that this gap represents a category-defining risk, and that solving it requires infrastructure-level intervention, not point solutions.

The identity crisis at the heart of agentic AI

Traditional cybersecurity anchors identity to humans, via multi-factor authentication, or to specific software services, via API keys and certificates. Autonomous agents break this model entirely. They are dynamic, often ephemeral, capable of spawning sub-agents, and able to operate across multiple systems in a single workflow; all without a consistent, verifiable identity that persists across those interactions.

According to Lyrie.ai founder Guy Sheetrit, the current ecosystem effectively allows agents to operate in anonymity. This creates three distinct high-risk scenarios: shadow agents accessing sensitive enterprise data without authentication; prompt injection attacks that manipulate an agent's decision logic to bypass permission sets; and attribution failure, where organisations cannot determine which autonomous process triggered a faulty or malicious transaction.

For enterprises in regulated GCC markets; where financial institutions operate under SAMA and UAE Central Bank frameworks, and where Saudi Arabia's and the UAE's data protection laws impose strict accountability requirements; attribution failure is not a technical inconvenience. It is a compliance liability.

Introducing the Agent Trust Protocol

The centrepiece of Lyrie.ai's launch is the Agent Trust Protocol (ATP), an open cryptographic framework designed to function as the identity and governance layer for autonomous AI agents; analogous to what TLS/SSL became for securing web communications.

ATP comprises three core mechanisms. Cryptographic authentication verifies an agent's identity fingerprint before it is permitted to interact with any system or data environment. Dynamic delegation controls govern how an agent can pass tasks to a sub-agent without losing the security context established at the outset of the workflow. And authorisation revocation allows enterprises to instantly terminate an agent's access if it exhibits anomalous behaviour or fails a real-time integrity check.

Critically, Lyrie.ai has submitted ATP to the Internet Engineering Task Force (IETF) for consideration as an open standard; a deliberate move to position the protocol as public infrastructure rather than a proprietary competitive moat. The approach mirrors the philosophy behind TLS: security at the infrastructure level works best when it is universal.

Offensive capability alongside defensive architecture

Lyrie.ai's platform does not stop at identity and access management. It integrates an offensive security layer; autonomous penetration testing that uses AI to find vulnerabilities in other AI agents before adversaries do, alongside attack simulations that stress-test agentic infrastructure against agent-jacking, logic manipulation, and exploitation of interfaces with legacy APIs.

This combination of identity infrastructure and adversarial testing reflects a sophisticated understanding of the threat landscape. Securing an agent's identity is necessary but not sufficient; enterprises also need to know whether the agents they deploy can withstand active attempts to subvert them.

Alignment with Anthropic and a path to Series A

Lyrie.ai's alignment with Anthropic's Cyber Verification Program signals a collaborative rather than purely competitive approach to AI safety. By working at the level of the foundational models that power today's most capable agents, Lyrie.ai ensures its security infrastructure is compatible with the platforms its enterprise clients are most likely deploying.

With $2 million in seed capital, the immediate roadmap focuses on scaling the cybersecurity research team with expertise in agentic vulnerabilities; a new class of security flaw unique to autonomous systems; advancing the ATP through the IETF approval process, and securing deployments in high-stakes enterprise environments where AI is used for critical infrastructure and financial operations.

The market timing is acute. Across the Gulf, agentic AI adoption is accelerating: 19% of GCC organisations have already moved from pilots to full-scale agentic AI implementation, with 74% planning adoption. As enterprises entrust autonomous agents with increasingly consequential actions, the question of whether those agents are who they claim to be; and whether they can be stopped when they are not; moves from a theoretical concern to an operational imperative.

Lyrie.ai's bet is that the answer to that question will require infrastructure-grade trust protocols. If ATP achieves the adoption trajectory of its TLS analogy, the startup's $2 million seed round may look like a very modest entry point into a very large category.