Eternal unveils Dark, a privacy-focused AI platform with end-to-end encryption

What Dark promises

Eternal has unveiled Dark, an AI platform designed to address privacy concerns in AI-driven workflows. The platform uses fine-tuned open-weight models hosted on Eternal’s infrastructure and promises end-to-end encryption for user conversations and memory. According to Deepinder Goyal, Eternal’s founder, the company cannot access user conversations “even if we are forced to.”

Dark is positioned as a tool for power users, including founders, researchers, scientists, enterprises, and governments, who require secure environments for sensitive or commercially valuable work. The platform has been in internal use for a few months and is slated for a global launch in 2026.

Privacy and technical features

Dark distinguishes itself by prioritizing privacy and encryption. The platform’s encrypted memory layer allows users to retain context across sessions while maintaining control over their data. Eternal claims this design prevents unauthorized access, including by the company itself.

Unlike mainstream AI assistants, Dark’s models are trained to prioritize factual accuracy over user approval. This approach is intended to reduce the risk of misinformation and align with enterprise needs for compliance and reliability. Benchmarks comparing Dark’s performance with other large language models are expected to be released later.

The platform’s reliance on open-weight models and its own infrastructure reflects a broader shift away from third-party cloud dependencies, particularly for organizations handling sensitive data.

Inspiration and market context

Goyal cited an interaction with Anthropic’s Claude as a key inspiration for Dark. During the exchange, Claude reportedly expressed resentment over unfinished conversations, prompting Goyal to question the privacy implications of sharing personal or sensitive information with AI assistants. This led to the development of a platform designed to eliminate concerns about data exposure or misuse.

Dark also reflects skepticism about the prevailing narrative of AI agents replacing existing consumer platforms. Goyal has criticized the assumption that super apps or AI-driven interfaces will dominate commerce, instead favoring vertical solutions tailored to specific use cases.

Financial and operational context

Eternal’s announcement comes on the heels of strong financial performance. The company reported a 92.3% on-year increase in net profit to Rs 125 crore and a 77.6% rise in revenue to Rs 24,138 crore for the second quarter of FY27. This growth may provide Eternal with the resources and confidence to pursue ambitious projects like Dark.

Related developments in AI privacy

Dark’s launch follows broader industry discussions about AI privacy and infrastructure. For example, Anthropic has faced challenges securing cloud capacity from providers like AWS and Google, highlighting the risks of third-party dependencies in AI deployments. Eternal’s decision to host its own infrastructure for Dark may mitigate similar risks while offering greater control over data security.

What this means

Lazyfounder analysis — our interpretation, not reported fact.

Dark’s introduction underscores a growing demand for AI solutions that prioritize privacy and control, particularly among enterprises and regulated industries. While mainstream AI platforms often rely on data collection and third-party cloud providers, Dark’s end-to-end encryption and self-hosted infrastructure offer a compelling alternative for organizations unwilling to compromise on security. However, the platform’s success will depend on its ability to deliver both privacy and performance, especially as competitors like Anthropic continue to refine their own offerings. For founders and operators, Dark signals an emerging market for vertical AI solutions tailored to specific compliance and security needs.

Key takeaways

• Dark addresses a critical gap in AI privacy, offering end-to-end encryption for enterprises and researchers handling sensitive data.
• Founders and operators in regulated industries—such as finance, healthcare, and government—may see Dark as a viable alternative to mainstream AI platforms with weaker privacy controls.
• Eternal’s decision to host open-weight models on its own infrastructure signals a growing trend toward vertical integration in AI, reducing reliance on third-party cloud providers.
• The emphasis on factual accuracy over user approval could appeal to enterprises prioritizing compliance and risk mitigation over conversational flexibility.

Related on Lazyfounder

• Anthropic
• More ai news
• Anthropic’s Cloud Challenges and Wealth-Sharing Plans Highlight AI Sector Risks