Deepinder Goyal, founder of Zomato and vice chairman of Eternal Ltd, has unveiled Dark.ai, a privacy-focused artificial intelligence chatbot designed to keep users’ conversations confidential through end-to-end encryption. The service is expected to launch globally by the end of 2026, positioning itself as an alternative for users who want to interact with AI without compromising sensitive personal or professional information.
Goyal revealed details of the upcoming platform in Eternal’s Q2 FY27 shareholder letter. He said Dark is being developed for people who want to discuss personal matters, explore confidential business ideas and conduct research without worrying about who might gain access to their conversations.
The chatbot has reportedly been used internally by Eternal’s team for several months. However, the company has yet to announce a confirmed public launch date or disclose details about its pricing and availability.
Credits: Digit
What Is Dark.ai and How Is It Different From ChatGPT?
Dark.ai is an AI chatbot designed to answer questions, help users explore ideas and provide assistance across a range of tasks. Its core distinction from established platforms such as ChatGPT and Google’s Gemini is its stated emphasis on privacy.
While existing AI services offer settings that allow users to manage how their conversations are used, disabling model training does not necessarily mean that conversations are never stored or accessed for other permitted purposes. Dark aims to address these concerns by using end-to-end encryption, which is intended to prevent anyone other than the communicating parties from accessing the underlying conversation.
According to Goyal, even the company operating the platform should not be able to read users’ chats, including in circumstances where it is compelled to provide access.
“The uncertainty and the lack of trust we have with large AI companies limits the power of AI in our lives,” Goyal wrote in the shareholder letter.
He believes privacy concerns can discourage people from asking AI assistants sensitive questions about health, finances and relationships. Similarly, founders and researchers may hesitate to share early-stage business concepts or confidential research if they are uncertain about how their information could be accessed.
Dark’s proposed privacy architecture is intended to address this trust gap and encourage users to make greater use of AI for sensitive tasks.
Dark.ai Features: Encrypted Conversations and Private Memory
Dark will run fine-tuned versions of advanced open-weight AI models on Eternal’s own infrastructure. Open-weight models allow developers to adapt existing models for particular applications without having to develop every component from scratch.
One of the platform’s central features is end-to-end encryption. Under this approach, conversation data is encrypted so that the service provider cannot read the messages in their ordinary form. The effectiveness of this protection will depend on how Dark implements encryption, manages keys and handles the interaction between users and AI models.
Another planned feature is encrypted memory, a user-controlled system intended to retain context across multiple sessions. This could allow the chatbot to remember relevant information and preferences without giving Eternal direct access to the underlying conversations.
The company is also working to improve the quality of its AI responses. Goyal said the models are being trained to prioritise what they believe is true rather than simply producing answers that please users. Eternal plans to publish benchmark results to provide further insight into the models’ performance.
However, the company’s claims about encryption and model quality will need to be evaluated against the technical details and independent assessments made available closer to launch.
Who Could Benefit From Dark.ai?
Dark is being developed with founders, researchers, scientists, businesses and government organisations in mind, particularly those that routinely handle confidential information.
For entrepreneurs, the platform could offer a way to brainstorm business strategies and explore product ideas with greater privacy. Researchers could potentially use it to examine early-stage hypotheses, while businesses may find encrypted AI tools useful when working with sensitive internal information.
The platform could also appeal to everyday users who want to discuss personal relationships, financial decisions or other private matters without exposing their conversations to the service provider.
Nevertheless, end-to-end encryption alone does not guarantee complete privacy across an entire service. Users will need clarity on how information is processed by AI models, whether external services are involved and how account data, backups and encrypted memory are managed.
Credits: Storyboard18
When Will Dark.ai Launch Globally?
Eternal is targeting a global launch for Dark by the end of 2026. The platform is currently undergoing internal use and testing, with further details about public access, pricing and supported features yet to be announced.
If its proposed encryption architecture delivers on its promises, Dark could offer a different approach to the growing AI chatbot market, where convenience and response quality have often taken priority over privacy.
For Goyal and Eternal, the challenge will be to combine strong confidentiality protections with reliable AI performance. The platform’s eventual success will depend not only on its privacy claims but also on whether users can independently verify those protections and trust the service with their most sensitive conversations.
