Protopia AI secures $1.25M Direct-to-Phase II SBIR contract from U.S. Air Force to deploy secure LLM coding infrastructure

Protopia AI, a Texas-based startup specializing in privacy-enhancing AI technology, has secured a $1.25 million Direct-to-Phase II Small Business Innovation Research (SBIR) contract from the U.S. Air Force to demonstrate its proprietary Stained Glass Transform (SGT) solution. The funding, issued via the Air Force Research Laboratory’s AFWERX innovation program in collaboration with the Air Force Sustainment Center Software Directorate (AFSC/SW), aims to evaluate SGT’s role in protecting sensitive codebases when using generative AI developer tools.

SGT, Protopia AI’s flagship product, is engineered to facilitate the secure use of proprietary or confidential data in machine learning environments. In this new project, the company will adapt SGT specifically for coding large language models (LLMs), ensuring that proprietary code can be used for inference and development workflows without being exposed in plaintext, even on shared or on-prem infrastructure.

The contract reflects rising defense-sector investment in trustworthy and mission-aligned AI adoption strategies. As software development within the Department of Defense increasingly involves tools like OpenAI’s Codex, Meta’s Code Llama, and other LLMs, new privacy-preserving methods are being prioritized to de-risk sensitive development environments.

Why is the U.S. Air Force investing in secure coding tools for generative AI?

This contract marks a critical pivot by the Department of Defense toward integrating cost-effective LLM-powered development workflows, while keeping operational security intact. The Air Force Sustainment Center Software Directorate, which oversees modernization of mission-critical maintenance and logistics software, will evaluate how SGT can enable widespread, secure adoption of coding LLMs by internal dev teams.

By applying SGT, the Air Force aims to strike a balance between AI-fueled productivity and preserving code confidentiality, especially in environments that depend on sensitive or classified codebases. Protopia’s solution addresses a core barrier in current LLM deployments—prompt leakage. When proprietary source code is submitted as a prompt to an LLM, it may be temporarily visible or stored in unprotected form, potentially exposing intellectual property, classified logic, or sensitive algorithms. SGT transforms this code into a format that is still usable by the LLM for inference while preventing plaintext exposure, thereby enabling privacy-first development acceleration.

Protopia AI will work closely with multiple AFSC/SW teams to operationalize SGT within their development pipelines, facilitating faster software delivery cycles with reduced cybersecurity risk. The platform is expected to help the Air Force unlock high-efficiency, on-premises generative AI infrastructure without forcing teams to trade off security for innovation.

Stained Glass Transform technology: origins and commercial relevance

Protopia AI’s Stained Glass Transform (SGT) is built on research originally incubated at the University of Texas at Austin. The company has previously demonstrated SGT’s effectiveness in securing machine learning inference pipelines across finance, healthcare, and edge AI use cases. With growing commercial interest in secure AI deployments—especially among financial institutions, biotech firms, and regulated software vendors—SGT is viewed as a viable approach to operationalize AI without compromising proprietary datasets.

Unlike traditional homomorphic encryption or secure multiparty computation approaches, which often incur prohibitive latency or compute costs, SGT uses an information-preserving transformation to mask sensitive input features while maintaining model performance. This makes it particularly suitable for inference-heavy tasks where latency and efficiency are key—such as with coding LLMs.

According to Protopia AI CEO and co-founder Eiman Ebrahimi, the contract not only validates the military utility of SGT but also positions it as a dual-use technology with strong commercial upside. “As AI-based productivity tools become central to value generation, it is essential to empower software development teams—both public and private—to use these models securely,” he stated.

SBIR funding as a bridge to dual-use commercialization

The $1.25 million Direct-to-Phase II award falls under the U.S. Air Force’s broader SBIR/STTR program, administered by the Air Force Research Laboratory (AFRL) and AFWERX. Direct-to-Phase II awards allow companies with existing prototypes or proofs-of-concept to skip early-stage feasibility and move straight into deployable development with federal funding.

For Protopia AI, this award serves both as a validation of its mature SGT platform and as an on-ramp to scale its solution for broader DoD and commercial integration. Dual-use technologies that satisfy military-grade performance and compliance standards are often viewed by VCs and strategic investors as lower-risk commercialization bets, particularly in today’s capital-constrained environment for deep tech.

The broader context includes recent increases in U.S. government funding toward “responsible AI,” a priority shared across the DoD, DHS, and NIST. The Defense Department’s Responsible AI Strategy and Implementation Pathway calls for “safe, secure, and effective” AI integration across all military functions—a mandate directly supported by Protopia’s SGT model.

Market reactions and expert sentiment on Protopia AI’s defense momentum

While Protopia AI is privately held and not yet publicly listed, investor and industry sentiment around the SBIR award suggests strong strategic positioning. The Direct-to-Phase II designation signifies that the company has already demonstrated technical feasibility and is prepared to scale solutions in production-grade defense environments—a stage where many startups struggle to gain traction.

Defense tech analysts note that software privacy has emerged as a bottleneck in LLM adoption, especially within federal agencies and defense contractor ecosystems. “Tools like Copilot and Codex offer great value but are typically excluded from secure environments due to prompt leakage risk,” said one independent consultant familiar with defense software procurement. “SGT effectively offers a way around that without compromising on throughput or accuracy.”

Early-stage institutional investors may also view this award as a signal of durability in Protopia AI’s business model. Government contracts, especially Direct-to-Phase II SBIRs, often act as non-dilutive capital infusions that reduce risk for future equity rounds.

Protopia’s positioning in the broader AI privacy market

Protopia AI joins a fast-evolving market of privacy-enhancing technology (PET) vendors, including names like Duality Technologies, Inpher, and Cape Privacy. However, its emphasis on usable, low-latency inference transformations—rather than privacy-preserving training or data synthesis—gives it a distinctive niche within the coding LLM and DevSecOps spaces.

Its work with AFSC/SW adds a key defense reference customer, positioning it favorably for future contracts with entities like the Defense Innovation Unit (DIU), the National Geospatial-Intelligence Agency (NGA), and other AI-forward arms of the federal government.

According to Crunchbase, Protopia AI had previously raised seed funding from institutional and strategic investors, but the company has not publicly disclosed a new financing round since this SBIR announcement. Still, the contract could serve as a strong proof point in the event of a Series A fundraising effort.

Analysts expect further AI security partnerships across the DoD

As generative AI tools continue to penetrate mission-critical workflows in government and industry, vendors offering practical solutions to the “LLM trust gap” are likely to attract sustained interest. With this SBIR award, Protopia AI enters a club of companies vetted for defense-grade privacy applications—a credential that could open doors across defense and critical infrastructure markets.

Looking ahead, analysts expect an uptick in procurement activity related to LLM security and AI DevOps tooling, as agencies push to adopt productivity-enhancing AI within their risk management boundaries. Partnerships between software infrastructure teams and PET innovators are forecast to increase, especially where cost-efficiency and on-prem AI compatibility are prioritized.

With Protopia AI now actively supporting AFSC/SW under a federally backed program, its Stained Glass Transform may find new applications beyond coding LLMs, including model serving for embedded systems, low-trust edge environments, and secure MLOps pipelines.