Telegram expands into confidential compute.

Telegram launched Cocoon on the TON network, its high-throughput blockchain, which was initially built to support large-scale messaging and payments.

Cocoon adds a confidential-compute layer that enables GPU providers to run parts of AI work that normally require access to user data, such as processing what you type, converting it into a format the model can understand, and generating the final answer, without exposing the information itself.

It matters because centralized AI providers typically have visibility into all of this data during processing, and Cocoon's design aims to limit that visibility.

This release aligns with a broader trend where more projects are migrating AI-related processes to blockchain-backed infrastructure to enhance privacy and data governance.

As AI begins to handle more sensitive data, confidential compute becomes a requirement in today's standards for data privacy and security.