Deepinder Goyal Teases Dark.ai, a Privacy-Focused AI Platform

Deepinder Goyal, founder of Eternal, has teased Dark.ai, a new AI platform focused on private conversations, with a global launch planned for 2026.

“Dark.ai – from the house of Eternal. Coming soon,” Goyal wrote in a post on X.

According to Eternal’s Q2 FY27 shareholders’ letter, Dark is built around the idea that users should be able to interact with AI without worrying about how their conversations are stored, accessed or used. 

The platform will use fine-tuned open-weight models running on Eternal’s own infrastructure. Dark is expected to launch globally in 2026.

“End-to-end encryption. We cannot read your conversations, even if we are forced to,” Goyal wrote in the letter, describing the platform’s approach to data privacy.

The service will include an encrypted memory management layer that allows context to carry across sessions while remaining under the user’s control. The company says conversations, research, ideas and other personal information will remain within an end-to-end encrypted environment.

“Dark is in the business of keeping you and your thoughts safe. Your thoughts, questions, insights, research, ideas, emotions, everything – stays within your end-to-end encrypted sandbox, so you can talk about anything without any lingering uncertainty in your mind,” Goyal said.

He further added that the product is intended for users who need to discuss sensitive information, including founders, researchers, scientists, enterprises and government organisations.

Goyal wrote that uncertainty over how AI companies retain and use user data can discourage people from sharing sensitive information with chatbots. Disabling model training on user data does not necessarily address other concerns, such as data retention, legal holds, abuse reviews, human audits and third-party logging, he said.

“The uncertainty and the lack of trust we have with large AI companies limits the power of AI in our lives,” Goyal said.

Dark will also use fine-tuned models that the company says are trained to prioritise truthful answers rather than simply please users. Eternal said it would reveal benchmarks for the models at a later date.

The company has been using the product internally for several months, according to Goyal. He said the experience had encouraged him to ask questions and explore research topics he would otherwise hesitate to discuss with mainstream AI assistants.