Eternal kündigt Dark.ai an: KI-Assistent mit Ende-zu-Ende-Verschlüsselung

Quick verdict: On Wednesday, Eternal — the company behind Zomato and Blinkit — told shareholders it's building Dark.ai, an AI assistant designed so Eternal itself cannot read your conversations. End-to-end encrypted, zero-access. Fine-tuned frontier open-weight models running on Eternal's own infrastructure, plus a user-controlled encrypted memory layer so the assistant keeps context across sessions without the company seeing it. Target: founders, researchers, scientists, enterprises, governments. Global launch "hopefully within 2026."

The honest version: this is an announcement, not a product. No launch date, no pricing, no demo, and nobody outside Eternal has audited the architecture. But the pitch is aimed at a real, specific gap — and Deepinder Goyal's one line is the entire thesis: "We cannot read your conversations, even if we are forced to." If that sentence survives contact with reality, it changes what AI is safe to use for. If it doesn't, it's another privacy press release. Here's both sides, straight.

The announcement came in Eternal's Q2 FY27 shareholders' letter, filed with India's BSE on October 21, 2026. Founder Deepinder Goyal wrote that Dark.ai has been in internal testing for several months — he says he now asks it things he wouldn't ask any other AI.

Three concrete pieces, straight from the letter:

That's the whole announcement. No model names, no benchmarks, no screenshots, no waitlist, no pricing.

How this was researched: built from Eternal's Q2 FY27 shareholders' letter and Oct 21–22 coverage from Business Standard, Economic Times, and Moneycontrol. Nothing here is hands-on testing — the product hasn't launched.

Here's the part most coverage will miss. The mainstream AI privacy debate is stuck on one toggle: don't train on my data. But opting out of training doesn't fix the rest of the list:

Dark.ai's claim is that the fix is architectural, not a setting: if the company never holds the keys, the list above mostly collapses. That's a serious argument. It's also currently unverified — architecture claims are proven by audits and cryptography, not shareholder letters.

Goyal names the audience precisely: "power users — founders, researchers, scientists, enterprises and governments — who want to keep their thoughts private, because privacy is essential for competitiveness." He adds that the models are fine-tuned "to not please you" — to tell the truth.

Read the list twice. Founders with unreleased products. Researchers with unpublished work. Governments. These are people whose downside isn't embarrassment — it's competitive or strategic damage. An AI you can't trust with secrets stays out of the rooms where the important work happens. "Privacy is essential for competitiveness" is the sharpest sentence in the letter, and it's the entire business case.

Every announcement gets a honeymoon. Here's the other side, in plain terms:

Dark.ai didn't land in a vacuum. The same week, Polarise launched a sovereign AI cloud at NVIDIA's GTC Berlin (Oct 22) — part of Europe's accelerating push for AI infrastructure it actually controls. Privacy and sovereignty are converging into one demand: AI you can use without handing your data — or your country's data — to someone else's stack.

Eternal is reading that room correctly. Whether a shareholder letter is enough to win it is a different question.

What is Dark.ai? Eternal's announced privacy-focused AI assistant (Q2 FY27 shareholders' letter, Oct 21, 2026): zero-access, end-to-end encrypted, fine-tuned frontier open-weight models on Eternal's own infrastructure, user-controlled encrypted memory. Not launched.

When will Dark.ai launch? No date. Global launch targeted "hopefully within 2026"; internal testing for several months per Goyal.

Who owns Eternal? Eternal Limited — parent of Zomato and Blinkit, founded by Deepinder Goyal.

Is Dark.ai really end-to-end encrypted? That's the announced design ("We cannot read your conversations, even if we are forced to") — but no independent audit or whitepaper exists yet, so it's unverified.

How is Dark.ai different from ChatGPT's "don't train on my data" setting? Opt-outs don't address retention, legal holds, abuse review, or third-party logging. Dark.ai claims an architectural fix: no keys, no access. That claim needs verification.

Is there pricing or a waitlist? Neither announced.

Sources: Business Standard, Economic Times, and Moneycontrol coverage, Oct 21–22, 2026.