Protopia AI
Austin-based AI security startup whose Stained Glass Transform technology obfuscates sensitive data before it reaches AI models, letting enterprises run inference on real data without exposing it.
Visit Website ↗ + Add to CompareOverview
Protopia AI, based in Austin, Texas, builds a data-protection layer for AI inference. Its core product, Stained Glass Transform, sits between an organization’s sensitive data and the AI or ML model consuming it, using learned transformations to obscure personally identifiable and confidential information so that a model can still extract useful signal without ever seeing the raw underlying data. A companion product, Stained Glass Engine, extends this to multi-tenant inference environments, letting shared AI infrastructure serve multiple customers or business units without hardware-level isolation.
The company was co-founded in 2020 by CEO Eiman Ebrahimi, a former NVIDIA research scientist, and Professor Hadi Esmaeilzadeh, an endowed chair in computer architecture at UC San Diego whose academic research underpins the core obfuscation technique. Protopia raised a $2 million pre-seed round followed by a $6 million seed round in December 2022, led by ATX Venture Partners with participation from Galaxy Interactive, Silverton Partners, and DNX Ventures, bringing total disclosed funding to roughly $8 million. The company has been beta-testing its technology with customers in financial services, government, and healthcare, where data-sharing restrictions have historically slowed AI adoption.
Protopia sits in a fast-growing but still-forming category: privacy-preserving AI infrastructure aimed at letting regulated enterprises use large models on sensitive data without violating privacy law or contractual data-handling obligations. Its differentiation is a mathematical transformation approach rather than the encryption-in-use or federated-learning techniques used by some competitors, but as an early-stage, seed-funded company its customer base and independent validation remain limited.
Innovation Matrix Assessment
Protopia has shipped two related products, Stained Glass Transform and Stained Glass Engine, since 2020, extending from single-tenant data obfuscation to multi-tenant inference isolation, a reasonable pace for a seed-stage team.
As a roughly 11-50 person seed-stage company still described as beta-testing with customers, Protopia has limited proven operational scale; it has not yet disclosed general-availability customer deployments at volume.
The company has raised a disclosed total of about $8 million (a $2M pre-seed plus a $6M seed round led by ATX Venture Partners), modest funding for the AI security space and no larger round has been reported since 2022.
Its approach, applying learned mathematical transformations to obscure sensitive data before it reaches a model, differs meaningfully from encryption-in-use, differential privacy, or federated-learning approaches used elsewhere in the privacy-preserving AI category.
The underlying technique traces to peer-reviewed academic research from co-founder Hadi Esmaeilzadeh's lab at UC San Diego, but the product itself is still in beta with financial services, government, and healthcare customers, so independent, named case studies were not found.
Enterprise AI adoption in regulated industries is frequently blocked by data-sharing and privacy rules, making inference-time data protection a highly relevant and timely problem as generative AI deployment accelerates.
Why CISOs Should Care
CISOs at regulated enterprises trying to let business units use AI on sensitive data without violating privacy law or data-handling contracts get a way to keep raw PII and confidential data out of model inputs entirely, rather than relying on downstream access controls alone.
What Makes It Different
Protopia's transformation-based approach obscures data before it reaches a model, rather than relying on hardware isolation, encryption-in-use, or federated learning, which is the more common approach among privacy-preserving AI competitors.
The Matrix Verdict
57/100 — INCREMENTAL INNOVATOR
Protopia AI addresses a genuinely important and growing problem, inference-time data privacy for enterprise AI, with academically-grounded technology, but it remains an early, thinly-funded, beta-stage company without independently verified production deployments.
Editorial Note: Claims vs. Verified Findings
The company's own descriptions of Stained Glass Transform's capabilities and its beta-customer industries (financial services, government, healthcare) are vendor-sourced and unverified by named case studies. The $6M seed round, its investor list, and the co-founders' backgrounds are independently corroborated by press coverage (VentureBeat, Built In Austin, Business Wire).
Sources
Alternatives to Protopia AI
Adaptive Security
AI-driven platform that simulates deepfake, voice, and multichannel social-engineering attacks to train and test organizations against next-generation phishing.
Quilr
Early-stage agentic AI security startup building a 'Service-as-Software' platform to guard against human-related breaches and secure AI agent…
Zenity
Governance and security platform for AI agents and low-code/no-code development, securing agent identity, permissions and behavior across the…
Tenzai
An agentic AI penetration testing startup building autonomous 'AI hackers' to find and validate exploitable vulnerabilities at a…
Reco
Reco secures the "agentic ecosystem" — mapping what AI agents can access across SaaS and enterprise apps, detecting…
Charm Security
Agentic AI workforce that investigates and intervenes on scams and fraud in real time, reading manipulation and intent…