Privatemode

The always-encrypted AI service.

Free Trial

Description

Privatemode offers a generative Artificial Intelligence service designed with data privacy and security as its core principles. It provides users with both a "Privatemode Chat" application for secure daily work, content generation, and document analysis, and a "Privatemode API" for developers to integrate privacy-preserving AI into their applications. Unlike many conventional AI services, Privatemode ensures that user data remains encrypted not only during transit and storage but also critically during processing. This is achieved through the use of confidential computing technology, preventing any unauthorized access to sensitive information.

The platform's security architecture is built on end-to-end encryption and end-to-end attestation, where the client application or proxy verifies the integrity of the service before any data transmission. This advanced zero-trust model ensures that even Edgeless Systems, the creators of Privatemode, cannot access user data. Privatemode allows the use of powerful open-source AI models, like Llama 3.3, with sensitive data, as the system is designed to prevent data leakage and use of data for model re-training. The service is hosted in the European Union, further underscoring its commitment to data protection standards.

Key Features

Confidential Computing: Encrypts data before it leaves the device and keeps it protected during AI processing.
End-to-end Encryption: Data remains encrypted during transfer, storage, and processing in main memory.
End-to-end Attestation: App and proxy verify the integrity of the service based on hardware-issued cryptographic certificates.
Advanced Zero-Trust Architecture: Prevents any external party, including Edgeless Systems, from accessing user data.
Private AI Chat: Secure AI assistant for daily work, content generation, and document analysis.
Private Inference API: On-premises-like privacy with cloud flexibility and scalability via an API key and proxy.
EU Hosting: Service hosted in top-tier data centers in the European Union.
No Data Training: Architecture ensures AI models cannot learn from or remember user data.

Use Cases

Secure content generation for businesses
Private analysis of sensitive documents
Developing AI applications requiring high data privacy
Using large language models without compromising confidentiality
Ensuring compliance in AI-assisted workflows

Frequently Asked Questions

What makes Privatemode more secure than other AI services?

Privatemode uses confidential computing to protect your data at all times, including during processing by the AI. This ensures no one, including Edgeless Systems or the cloud provider, can access your data, offering a higher level of security than contractual agreements alone.

Which AI models does Privatemode provide?

Privatemode currently uses a 'quantized' version of Meta's Llama v3.3 with 70B parameters. Other models like DeepSeek R1 are planned to be added soon.

Will the AI learn from my data or remember it?

No. Privatemode's architecture, based on confidential computing, ensures that the employed AI model can technically not learn from your data or remember it in any form.

How can the AI process my data if my data is 'always encrypted'?

Privatemode uses confidential computing, where the CPU and GPU hardware selectively decrypt your data into internal, inaccessible structures for processing. From the outside, it appears as if the hardware processes encrypted data, protecting it even from privileged access.

Why do I need to download an app to use Privatemode Chat?

The Privatemode app is required to verify the integrity of the Privatemode service using remote attestation and to encrypt all data before transmission. These security tasks cannot be performed as efficiently or securely within a standard web browser.