Eternal’s Dark.ai aims to offer encrypted AI chats, targets launch this year

Eternal is preparing to launch Dark.ai, an artificial intelligence platform focused on privacy and encrypted conversations. The platform is expected to launch globally this year, according to details shared about the initiative.

Dark.ai aims to address concerns around data privacy in mainstream AI assistants, where users may hesitate to discuss sensitive topics or share unpublished research due to uncertainty over how their conversations are stored and used.

The platform's introduction highlights concerns around data retention, legal requests, abuse reviews and third party logging systems. It argues that disabling model training on user data does not necessarily address these concerns.

Dark uses fine tuned frontier open weight models that run on its own infrastructure. The platform claims to offer end to end encryption, with the company stating that it cannot read users’ conversations, even if compelled to do so.

The platform also features an encrypted memory management layer that allows context to carry across sessions under the user's control. It is designed for power users, including founders, researchers, scientists, enterprises and governments, who want to keep their ideas and conversations private.

The platform's introduction also highlights how privacy concerns can limit the use of AI tools. Researchers may hesitate to share early stage ideas, while users may avoid asking sensitive health, financial or relationship questions. Dark aims to provide an environment where users can discuss such topics without fear of judgment or exposure.

Beyond privacy, the platform aims to offer answers that prioritise factual accuracy over user approval. Its models are trained to tell users what they think is the truth rather than simply please them. Benchmarks comparing their performance with other large language models (LLMs) are expected to be revealed later.

Dark is being developed for users who require privacy as well as high quality AI responses. However, details about its pricing, commercial model and the technical implementation of its encryption system remain undisclosed.

The platform is expected to launch globally this year, according to the information shared about the initiative.

Show reader opinions