Private AI was a Canadian data-privacy company that built a PII detection and redaction platform for unstructured text. Its 'Private GPT' product (accessible at private-ai.com/private-chatgpt/) sits as a proxy layer between users and ChatGPT: it strips personally identifiable information, protected health information, and payment card data from prompts before they reach OpenAI's servers, then re-populates the redacted values in the returned answer so the interaction remains seamless. In early 2026 (around March 2026), Private AI rebranded as Limina, deriving the name from the Latin word for 'threshold'. The rebrand did not change the company's products, customers, or operations; the Private GPT product continued to operate under the new Limina brand, with the customer portal at portal.getlimina.ai/privategpt. The core technology detects 50+ entity types across 52 languages with accuracy the company claims surpasses human review. General-purpose solutions miss between 13.8% and 46.5% of PII entities in real-world datasets, while Limina's system misses between 0.2% and 7% across the same benchmarks, according to the company's own published comparisons. The platform handles multiple file types including PDFs, audio files, and images through a single API. The target market is regulated industries that need to interact with cloud LLMs without exposing sensitive data: healthcare, life sciences, insurance, and financial services. The solution addresses compliance requirements including GDPR, HIPAA, and CCPA by ensuring raw sensitive data never reaches an external model provider. Deployment options include both cloud and on-premises to accommodate data residency requirements. Note: This product (by private-ai.com, now getlimina.ai) is distinct from the open-source zylon-ai/private-gpt GitHub project, which is a self-hosted document-chat application maintained by Zylon. The two share a similar name but are entirely separate products from different companies. Key features: - PII/PHI/PCI redaction across 50+ entity types before prompts reach ChatGPT - Automatic re-injection of redacted values into ChatGPT responses for seamless UX - Support for 52 languages - Handles text, PDFs, audio files, and images via a single API - Cloud and on-premises deployment options for data residency compliance - GDPR, HIPAA, and CCPA compliance layer - Higher-than-human accuracy with claimed miss rate of 0.2%-7% vs 13.8%-46.5% for general-purpose tools
Enterprise pricing starts at approximately $39k per year, scaling to $400k+ per year before bespoke enterprise contracts. No free tier or self-serve trial publicly listed.