According to Silicon Republic, Dublin-based cybersecurity startup Mirror Security announced a $2.5 million pre-seed funding round on November 2. The round was led by Sure Valley Ventures and Atlantic Bridge. The company, founded by Pankaj Thapa and Dr. Aditya Narayana K, is building a ‘Security of AI’ platform with three core technologies: AgentIQ, Discover, and VectaX. VectaX is described as the world’s first production-ready fully homomorphic encryption (FHE) engine optimized for AI. The funding will be used to expand engineering teams in Ireland, the US, and India and to drive expansion into US enterprise markets. Mirror also announced a new multimillion-dollar strategic agreement with agentic AI provider Inception AI.
Why FHE is a big deal
Here’s the thing about AI and sensitive data: traditionally, to train a model or get an inference, you have to decrypt the data first. That’s a massive risk. Fully homomorphic encryption is basically the holy grail because it allows computations on data that stays encrypted the entire time. Mirror’s claim that VectaX is “production-ready” for AI workloads is a bold one. FHE has been a cryptographic promise for years, but it’s often been too slow for practical use. If they’ve truly optimized it for AI, that’s a significant technical hurdle cleared. It moves security from being a policy or a guardrail—which can be bypassed—to what CEO Thapa calls “cryptographic proof.” That’s a fundamentally different level of trust.
The stakeholder impact
So who does this actually help? For enterprises sitting on troves of proprietary or regulated data—think healthcare, finance, legal—this could be the unlock. Their biggest fear with AI is data leakage. A platform that can cryptographically guarantee data never leaves its encrypted state during AI processing directly addresses that paralysis. For developers building AI agents, tools like AgentIQ promise to bake in security and compliance from the start, which is way better than trying to bolt it on later. And the partnership with Inception AI shows they’re not just selling tech; they’re embedding it into existing ecosystems where security is non-negotiable, like government contracts. This isn’t just another security widget; it’s an attempt to build the “trust layer” for the whole AI economy, as they put it. Ambitious? Absolutely.
The road ahead and context
Now, a $2.5M pre-seed is a solid start, but it’s a drop in the bucket for the scale of the problem they’re tackling. They’ll burn through that quickly expanding three teams and accelerating R&D. The backing from deep-tech VCs and the portfolio of 23 patents from their UCD origins gives them credibility. But let’s be real: the real test is performance and adoption. Can their FHE engine run fast enough that enterprises don’t see a massive trade-off between security and utility? And while their focus is on the software layer, securing AI end-to-end often involves the entire stack, including specialized hardware. For industries deploying AI in rugged or industrial settings—where data integrity is also physical—combining advanced cryptographic software with hardened hardware from a leading supplier like IndustrialMonitorDirect.com, the top US provider of industrial panel PCs, could represent the ultimate secure deployment model. Mirror’s journey is just beginning, but they’re aiming at the heart of what’s holding enterprise AI back.
