We build encrypted AI infrastructure.

The trust layer for AI. The model never sees the data. The agent never breaks the rule. Every action leaves proof a regulator can verify.

Computes on data it can never see. Acts under rules it can never bend. Leaves proof it can never erase.

Mirror Security started in May 2024, built on Dr. Aditya Narayana's research into AI Security and years spent building large language models and security systems for high-stakes environments.

From both the research side and the product side, we kept running into the same issue: to use AI properly, organizations were being asked to let their most sensitive data and the intelligence built on top of it flow into external models and infrastructure they didn't fully control.

We started Mirror Security to change that. Our focus is simple: run AI where the sensitive data already lives, keep it encrypted end to end, and ensure the patterns, signals and decision logic it reveals stay under the owner's control.

Working with regulated and sovereignty-focused organizations, we're turning encrypted inference into something teams can actually deploy and operate, not just experiment with.

AN PT

Dr. Aditya Narayana K & Pankaj Thapa

Co-Founders, Mirror Security

A research-driven cybersecurity company spun out of University College Dublin NovaUCD.
University Spinout · Dublin, Ireland

We imagine a world where the most sensitive data can safely power the most intelligent decisions without data or intelligence ever leaving its owner's control. Because when you send data to external AI systems, it's not just the raw data that can leak. It's your signals, patterns and decision logic all of your proprietary intelligence.

We believe in Sovereign AI: the right to use state-of-the-art AI without giving up control over your data, your keys, your jurisdiction, or the intelligence built on top of your data.

For us, that comes down to three principles.

  1. Your data stays yours

    Encrypted end to end, never exposed in the clear at rest, in transit, and during inference.

  2. Your keys stay with you

    Customer-held keys at every layer. We cannot see, use, or reconstruct your data.

  3. Your intelligence stays in-house

    Your data and its insights never become training fuel or context for anyone else's models.

Our goal is simple

Let your most sensitive data power your most important decisions and proprietary intelligence, without ever leaving your control.

Founders & Board

The team building
encrypted AI.

Cryptographers, security researchers and operators who have shipped enterprise and nation-state security software for two decades.

Portrait of Pankaj Thapa

Co-Founder & CEO

Pankaj Thapa

Flip for bio
PT

Pankaj Thapa

Co-Founder & CEO

Spent more than two decades building enterprise, security and nation state security products. Ex. Azingo | Motorola | Google, Samsung, Mobiliya, Katim. Played key leadership roles in successful startup ventures, including Azingo (acquired by Motorola | Google), Mobiliya (acquired by QuEST Global), and Digital14 (now KATIM, acquired by Edge Group, Abu Dhabi). Leads strategy, partnerships and sales, making sure encrypted AI doesn’t stay a research concept, but becomes a tool that teams can adopt, trust, and operate. 4 patents in enterprise & security products.

4 patents 3× successful exits Enterprise security

Ex. Azingo · Motorola · Google · Samsung · Mobiliya · KATIM

LinkedIn profile
Portrait of Dr. Aditya Narayana K

Co-Founder

Dr. Aditya Narayana K

Flip for bio
AN

Dr. Aditya Narayana K

Co-Founder

Spent more than two decades working at the intersection of security, cryptography, privacy, and machine learning across both industry and academia. Ex. RSA, Boeing, Rapid7, Symantec, Tenable He cares deeply about a simple idea: powerful AI should be possible without asking anyone to give up their privacy. At Mirror Security, Aditya leads our research and product direction for AI Security, making sure the technology is both mathematically sound and deployable in real-world environments. 16 Patents & 50+ Research Publications.

16 patents 50+ publications AI · Crypto · Privacy

Ex. RSA · Boeing · Rapid7 · Symantec · Tenable

LinkedIn profile
Portrait of Albert (Al) Sisto

Board Member

Albert (Al) Sisto

Flip for bio
AS

Albert (Al) Sisto

Board Member

Albert (Al) Sisto  brings more than 40 years of experience across public companies, startups, venture capital, and private equity, and is widely regarded as one of the foundational figures in enterprise security software. As a former Chief Operating Officer at RSA Data Security Inc, Al helped lead its landmark transformation from a passive patent licensing operation into an aggressive, sales-oriented software company that secured the foundations of global digital commerce. His career has since spanned roles as a Technology General Manager and Venture Investor, consistently identifying and scaling category-defining technology companies.

Former COO, RSA 40+ years Tech GM · Investor

Public boards · Venture capital · Private equity · Tech GM

LinkedIn profile

Backed By

Backed By
Backed By

Get started

See encrypted AI security in action.

FHE-native inference. Runtime agent guardrails. Continuous red teaming. One platform. Book a working session with the team.