Liminal

Liminal is a comprehensive security and governance platform built for enterprises to harness the power of generative AI without compromising data security or compliance. Trusted by leading organizations in regulated industries, Liminal acts as a secure intermediary between employees and various large language models (LLMs) like GPT-4o, Claude 3, and Gemini. The platform is designed to empower the entire workforce to boost productivity with AI, while providing security and privacy teams with the robust controls they need for complete oversight.

The core of Liminal is its powerful data protection engine, which uses a proprietary ensemble algorithm to detect, protect, and rehydrate sensitive information. This ensures that confidential data such as Personally Identifiable Information (PII), Protected Health Information (PHI), intellectual property, and trade secrets never leave the enterprise's secure environment. By intelligently masking or redacting data before it reaches an LLM and then restoring it in the response, Liminal maintains the context of the user's query, delivering accurate results without data exposure.

How to use Liminal

Getting started with Liminal is designed to be quick and straightforward, often taking less than an hour to get up and running. The process typically involves the following steps:

1. Schedule a Demo: The first step is to contact the Liminal team to schedule a personalized demo and discuss specific organizational needs.
2. Deployment: Liminal can be deployed quickly within your enterprise environment. It supports single-tenant architecture for maximum security.
3. Configuration: Security and IT administrators configure the platform through a central hub. This includes setting up granular access controls, defining data protection policies (e.g., what data to mask, redact, or block), and integrating with existing Identity Providers (IdP) and SIEMs for logging.
4. User Onboarding: Employees can then start using generative AI through Liminal's various tools. This includes the Secure Multi-Model Chat, the Liminal Helper (which integrates into any website, web app, or desktop platform), and custom-built Model-Agnostic Assistants.
5. Secure Interaction: As users submit prompts, Liminal's engine automatically intercepts them, applies the configured security policies to protect sensitive data, sends the sanitized prompt to the selected LLM, and rehydrates the response for a seamless user experience.
6. Monitoring and Governance: Security teams have real-time observability into all AI usage across the organization, with detailed logs, analytics, and alerting capabilities to ensure compliance and prevent misuse.

Core Features of Liminal

• Secure Multi-Model Gateway: Provides unlimited, secure access to a wide range of leading LLMs, including GPT-4o, Claude 3.7 Sonnet, Gemini 1.5, and even internal models.
• Advanced Data Protection: Automatically detects and protects sensitive data (PII, PHI, PCI, IP) in prompts using intelligent masking and redaction, preserving context for better AI responses.
• Seamless Workflow Integration (Liminal Helper): A context-aware tool that brings secure generative AI into any application employees use, including Microsoft 365, Google Workspace, Salesforce, and Slack.
• Custom Model-Agnostic Assistants: Allows organizations to build, train, and share their own secure, custom AI assistants tailored to specific tasks or departments, without using sensitive data for training.
• Comprehensive Administration and Observability: A single hub for security teams to manage policies, control access by role or model, and monitor all AI interactions with real-time logging and alerting.
• Developer SDK (Liminal Elements): An SDK that allows developers to easily embed Liminal's security controls into custom-built employee or customer-facing applications with just a few lines of code.
• Enterprise-Grade Compliance: The platform is HIPAA compliant, SOC 2 Type 1 & 2 certified, and is working towards ISO 27001, HITRUST, and NIST certifications, making it ideal for regulated industries.

Use Cases for Liminal

Liminal is designed for a wide range of industries where data security is paramount:

• Healthcare & Life Sciences: Clinicians and researchers can use AI to analyze patient data, summarize medical records, and accelerate research without exposing Protected Health Information (PHI).
• Financial Services, Banking & Credit Unions: Financial professionals can leverage AI for market analysis, fraud detection, customer service, and report generation while ensuring compliance and protecting sensitive financial data.
• Insurance: Teams can streamline claims processing, perform risk assessments, and personalize policies using AI, with full confidence that customer data is secure.
• Education: Educational institutions can use generative AI for curriculum development, student support, and administrative tasks while safeguarding student privacy.
• Legal and Professional Services: Lawyers and consultants can use AI for case research and document review without risking client confidentiality.

Advantages of Liminal

Liminal offers a unique combination of robust security and a user-friendly experience. Key advantages include preventing data leakage to public AI models, ensuring regulatory compliance, eliminating the risk of 'Shadow AI' by providing a sanctioned and easy-to-use tool, and boosting workforce productivity by making powerful AI accessible everywhere. Its model-agnostic approach provides flexibility and future-proofs the organization's AI strategy. The focus on intelligent masking and rehydration ensures that security measures don't hinder the quality of AI outputs or disrupt user workflows.

Pricing and Plans

Liminal's pricing information is not publicly listed on their website. As an enterprise-focused platform, they operate on a custom pricing model tailored to the specific needs of an organization, such as the number of users, required features, and deployment complexity. To get pricing details, interested parties are encouraged to contact their sales team or schedule a demo through the official website.