Lattica
Lattica enables AI inference on encrypted data without exposing sensitive information.
Lattica is a privacy-preserving AI inference platform built on Fully Homomorphic Encryption (FHE) that enables organizations to run AI models on encrypted data without exposing sensitive information to model providers, cloud infrastructure, or intermediaries. The platform features a hardware-agnostic Homomorphic Encryption Abstraction Layer (HEAL) that optimizes FHE algorithms to run in parallel on GPUs, solving the historical CPU-bottleneck problem. It integrates seamlessly with PyTorch and standard ML frameworks, allowing developers to build privacy-preserving applications without cryptography expertise. Lattica targets enterprises in regulated industries handling sensitive data who need confidential AI inference in cloud environments.
Problem solved
Organizations cannot safely deploy AI models on sensitive data because cloud providers and model vendors gain access to raw inputs, creating compliance, privacy, and competitive risk.
Target customer
Enterprise organizations in regulated industries (healthcare, finance, government) requiring confidential AI processing in cloud environments; AI model providers seeking to offer privacy-preserving inference capabilities.
Founders
D
Dr. Rotem Tsabary
Founder & CEO
PhD in lattice-based cryptography from Weizmann Institute of Science; previously engineer at Google working on AI inference acceleration.
Funding history
Pre-Seed
$3.25M
April 23, 2025
Led by Cyber.Fund (Konstantin Lomashuk)
· Sandeep Nailwal (Polygon, Sentient), Lightshift Capital, XT Venture Capital
Total raised:
$3.25M
Pricing
Not publicly available. Company is in early-stage deployment phase; pricing likely to be disclosed as product matures and customer base expands.
Notable customers
Not disclosed. Company emerged from stealth in April 2025 and has not publicly announced customer names or case studies.
Integrations
PyTorch, NVIDIA GPUs, hardware-agnostic support for TPUs, CPUs, ASICs, FPGAs
Tech stack
WOW (JavaScript frameworks)
jQuery Migrate (JavaScript libraries)
jQuery (JavaScript libraries)
FancyBox (JavaScript libraries)
Bootstrap Icons (Font scripts)
Popper
RSS
Open Graph
HTTP/3
WordPress (Blogs)
Google Analytics (Analytics)
Cloudflare Bot Management (Security)
Twitter Emoji (Twemoji)
PHP (Programming languages)
Google Workspace (Email)
Cloudflare (CDN)
MySQL (Analytics)
Google Tag Manager (Tag managers)
Yoast SEO Premium (SEO)
Yoast SEO (SEO)
Kinsta (PaaS)
CookieYes (Cookie compliance)
Contact Form 7 (WordPress plugins)
Website
Competitors
Fhenix
Blockchain-focused FHE platform for confidential DeFi applications on Ethereum, rather than general-purpose cloud AI inference.
Inco
Web3-oriented confidentiality layer emphasizing blockchain integration, whereas Lattica focuses on practical enterprise AI inference in cloud.
Synnax Technologies
Specializes in decentralized AI-powered credit intelligence with FHE; narrower industry focus compared to Lattica's general-purpose AI inference platform.
Primus
Focuses on zkTLS and zkFHE interoperability layer; different cryptographic approach and narrower use case compared to Lattica's comprehensive FHE inference platform.
Why this matters: Lattica solves a critical bottleneck in privacy-preserving AI by successfully parallelizing Fully Homomorphic Encryption on GPUs—a problem that has plagued the field for years. Its emergence from stealth with significant backing from prominent crypto investors signals growing enterprise demand for cryptographically-proven confidential computing, positioning it at the intersection of AI, security, and regulated industry requirements.
Best for: Enterprises in healthcare, finance, and government sectors that need to run confidential AI inference on sensitive data while maintaining compliance with privacy regulations and protecting proprietary information.
Use cases
Healthcare AI diagnosis on patient data
A healthcare provider wants to use AI models for patient diagnosis without exposing protected health information (PHI) to the cloud provider or model vendor. Lattica encrypts patient data in the browser, processes it through AI models while encrypted, and returns only the diagnosis result—ensuring HIPAA compliance and patient privacy.
Financial institution credit analysis
A bank needs to run proprietary credit scoring models on customer financial data without exposing either customer information or proprietary model weights to cloud infrastructure. Lattica enables encrypted inference so only credit decisions are revealed, protecting both customer privacy and competitive model advantage.
Government intelligence analysis
A government agency requires classified data processing through AI models where neither the cloud infrastructure nor model providers see raw intelligence inputs. Lattica's FHE approach ensures classified information remains encrypted throughout inference, meeting security clearance and compartmentalization requirements.
Alternatives
Trusted Execution Environments (TEEs)
TEE solutions (Intel SGX, ARM TrustZone) rely on hardware trust assumptions; Lattica uses cryptographic guarantees that work across any hardware without requiring trust in cloud providers.
Differential Privacy
Differential privacy adds statistical noise to protect individual records; Lattica provides deterministic encryption of data, offering stronger privacy guarantees for inference without accuracy loss.
Secure Multi-Party Computation (MPC)
MPC requires coordination between multiple parties and is computationally expensive for large-scale inference; Lattica's FHE enables single-party encrypted inference at scale on GPUs.
FAQ
What does Lattica do? +
Lattica is a platform that enables AI models to process encrypted data without any party (including Lattica, cloud providers, or model vendors) seeing the raw information. It uses Fully Homomorphic Encryption optimized to run in parallel on GPUs, and includes a hardware-agnostic abstraction layer (HEAL) that integrates with PyTorch and standard ML frameworks.
How much does Lattica cost? +
Pricing is not publicly available. The company emerged from stealth in April 2025 and is in early deployment stages. Contact sales for enterprise pricing information.
What are alternatives to Lattica? +
Alternatives include Trusted Execution Environments (TEEs) like Intel SGX which rely on hardware trust; differential privacy solutions that add statistical noise for privacy; and Secure Multi-Party Computation (MPC) which requires coordination between multiple parties. Lattica differs by using cryptographic encryption that provides zero-knowledge guarantees across any hardware.
Who uses Lattica? +
Target customers are enterprises in regulated industries (healthcare, finance, government) that need confidential AI inference. As of April 2025, specific customer names have not been publicly disclosed; the company is in early deployment phase.
How does Lattica compare to Fhenix? +
Fhenix focuses on blockchain-based encrypted computation for DeFi applications on Ethereum, whereas Lattica targets general-purpose enterprise AI inference in cloud environments. Lattica's GPU-optimized FHE implementation is designed for practical, scalable AI model serving in non-blockchain contexts, making it more suitable for traditional enterprise use cases.
Tags
fully homomorphic encryption
privacy-preserving AI
confidential computing
encrypted inference
GPU acceleration
machine learning
data privacy