ExpressVPN has launched ExpressAI, a private AI platform it says is designed to let users access AI tools without their prompts and files being used for model training.
The company said ExpressAI uses confidential computing “secure enclave” technology so that conversations are decrypted only inside a cryptographically isolated environment intended to be inaccessible to the host system and infrastructure operators. ExpressVPN said this is designed to keep user data encrypted and unreadable, including to ExpressVPN itself.
ExpressAI includes a selection of third-party models at launch, including options from OpenAI and DeepSeek, as well as Qwen and NVIDIA’s Nemotron. ExpressVPN said the service supports a side-by-side comparison view to run the same prompt across multiple models.
In the announcement, ExpressVPN said ExpressAI includes “zero-access” encryption and a “zero training” approach, meaning user conversations, prompts and files are not fed back into model training. Other features listed include private file handling, an auto-delete “ghost mode”, and an “encrypted vault” where users set a password to decrypt chat history.
ExpressVPN said ExpressAI has been independently audited by security assessment firm Cure53. The audit, conducted in February and March 2026, included penetration testing and source code review across the platform’s front-end and back-end systems, cryptography, key management, and infrastructure. ExpressVPN said findings were remediated ahead of launch and linked to the published report.
“The best way to protect user data is not to collect it in the first place. We’re not just making privacy claims – we’re proving it with cryptographic guarantees. With our enclave architecture, your messages exist in a secure, isolated environment that even we can’t access. This is the future of reliable, private AI,” said ExpressVPN COO Shay Peretz.
ExpressVPN said ExpressAI will roll out in phases, with new and existing ExpressVPN Pro plan users receiving first access.
