X

xAI Chatbot Grok Will Be Available to Telegram Users

Billion-user-strong messenger app will begin integrating Elon Musk's AI chatbot this summer

Headshot of Alex Valdes
Headshot of Alex Valdes
Alex Valdes
Alex Valdes from Bellevue, Washington has been pumping content into the Internet river for quite a while, including stints at MSNBC.com, MSN, Bing, MoneyTalksNews, Tipico and more. He admits to being somewhat fascinated by the Cambridge coffee webcam back in the Roaring '90s.
Alex Valdes
2 min read
grok ai logo

Telegram CEO Pavel Durov announced on X that he and Elon Musk have agreed on a one-year deal to bring xAI's Grok to the Telegram.

 Viva Tung/CNET

Elon Musk's xAI chatbot Grok will be available to Telegram users for one year starting sometime in the summer under a deal announced Wednesday. Telegram is a cloud-based social media and instant messaging service that is free but offers a paid tier.

Telegram CEO Pavel Durov, who announced the partnership on X, said Telegram users will be able to pin Grok to the top of their chats. Users will be able to ask questions, create stickers, write suggestions and summarize chats, links and documents.

An X user asked Durov about the security of user data.

"User privacy is paramount," Durov said. "To be clear, xAI will only access data that Telegram users explicitly share with Grok through direct interactions. That's expected -- you can't message anyone (including a chatbot) without sharing what you write."

Telegram, with an estimated 1 billion active users, will receive $300 million in cash and equity from xAI to integrate Grok into Telegram apps for 12 months. Durov's company will also receive half of the revenue from xAI subscriptions purchased through Telegram apps.

Telegram premium users have had access to Grok since early 2025.

Until now, X has trained Grok's responses through public posts, except in the EU. It is unclear whether the Telegram-xAI partnership will leverage Telegram user posts to further train Grok.

With AI chatbots still relatively in their infancy, there are still plenty of issues to iron out. A recent study out of Israel found that it is relatively easy to trick chatbots into giving out answers that are supposed to be filtered out, such as violent videos or information not intended for public eyes.

Another study from Microsoft and Salesforce found that AI chatbot accuracy faltered significantly the longer the conversation went. Models such as Claude 3.7 Sonnet, Gemini 2.5 Pro and GPT-4.1 had a 30% to 40% drop in accuracy in "multi-round conversations," according to the research.

For an in-depth primer on AI chatbots, check out CNET's What's an AI Chatbot? Everything to Know.