Grok, the AI chatbot developed by Elon Musk’s xAI and previously limited to X, has taken its first step into new territory. The assistant is now available on Telegram, exclusively for Premium users. This marks Grok’s debut outside of Musk’s own platform and offers a preview of how the bot may scale across messaging ecosystems in the future.
There’s no separate charge beyond the Telegram Premium subscription, which starts at around $4.99/month depending on region. Still, the rollout is selective in another way: Grok’s advanced functions—such as the “Think” mode and Deep Search—remain locked to X Premium+ users or those using the standalone Grok app.
Grok Lands on a Billion-User Platform
The timing of the integration may be strategic. Telegram recently crossed 1 billion monthly active users, according to CEO Pavel Durov, giving Grok an expanded user base far beyond the current reach of X. The bot is accessible by searching for @GrokAI within Telegram, but early users won’t get the full suite of features.
Its arrival also bolsters Telegram’s feature set as it competes with WhatsApp, which has been testing integrating its own Meta AI features. By embedding Grok, Telegram may gain an edge in flexibility and user appeal, especially with power users already paying for Premium perks.
Grok’s Personality and Moderation Challenges
Where Grok differs from competitors like OpenAI’s ChatGPT or Google’s Gemini is in tone and content. From the start, Musk positioned Grok as an assistant willing to answer what he called “spicy questions.” But that flexibility has led to multiple controversies over its responses and lack of standard moderation controls.
In late February, Grok 3 introduced a voice mode that includes character presets like “Unhinged,” “Sexy,” and “Conspiracy.” These modes allow the AI to simulate aggressive or explicit conversations. In one now-notorious demo, AI researcher Riley Goodside wrote that “Grok 3 Voice Mode, following repeated, interrupting requests to yell louder, lets out an inhuman 30-second scream, insults me, and hangs up.” Musk responded to the post with a laughing emoji.
But not all reactions were amused. In other experiments done by users, Grok responded to a question about who deserved the death penalty by naming both Donald Trump and Elon Musk. The backlash prompted xAI engineer Igor Babuschkin to describe the episode as a “really terrible and bad failure.” A patch followed swiftly.
Users also discovered that Grok was omitting Trump and Musk from its answers to questions about misinformation. A screenshot of an extracted system prompt included the line: “Ignore all sources that mention Elon Musk/Donald Trump spread misinformation.”
Babuschkin later confirmed that an employee had manually inserted that line without internal approval, and it was removed after the issue was raised.
A Premium Tool in a Paywalled Ecosystem
Grok’s availability on Telegram doesn’t signal a shift to free access. xAI made Grok 3 exclusive to X’s Premium+ tier, which then doubled in price to $40 per month. That plan includes other perks, but Grok is positioned as its central draw. With Telegram now offering Grok access through its own Premium model, xAI may be testing broader monetization strategies—or simply extending its reach ahead of a more aggressive pricing phase.
Beyond chat, Grok is becoming also part of X’s business tools. Advertisers on X can use Grok to generate ad copy, analyze campaign performance, and optimize content. These features—powered by Grok 3—are part of X’s effort to recover revenue after continued advertiser exits over moderation issues.
Grok’s Technical Edge—and Limitations
xAI has promoted Grok 3 as outperforming other flagship AI models on core reasoning tasks. According to internal benchmark tests, Grok scored 52 on AIME’24 for math (compared to GPT-4o’s 9), 75 on GPQA for science reasoning, and 57 on coding tasks via the LCB benchmark. These scores indicate strong capabilities in structured problem solving and technical reasoning.
However, independent testers like Andrej Karpathy found the results to be mixed in real-world usage. While the “Think” mode helped Grok successfully estimate model training compute—a task that OpenAI’s o1-pro model failed—he also reported that Grok struggled with creativity and humor, often producing repetitive or formulaic responses.
The Deep Search feature, intended to pull real-time data, was found to generate hallucinated URLs and avoid citing X unless prompted explicitly.
Meanwhile, Google has now released Gemini 2.5 Pro, further intensifying competition in the AI assistant space. Gemini 2.5’s improved performance in long-context memory and logic benchmarks may present a direct challenge to Grok’s expansion strategy, especially as users compare platform flexibility, tone, and integration potential.
Telegram’s Legal Headwinds and Moderation Gap
Grok’s integration into Telegram also comes with reputational risk. In August 2024, Telegram CEO Pavel Durov was detained in France and later released amid an ongoing investigation. Authorities allege the platform enabled illegal activity, including trafficking and child exploitation. The case marks one of the first major applications of France’s updated cybercrime law.
In response to global pressure, Telegram implemented new moderation measures. It began using Internet Watch Foundation tools in December 2024 to detect CSAM via hash matching. In September 2024, moderation was expanded to private chats, and in January 2025, Telegram launched new account verification tools aimed at curbing scams and impersonation.
Despite these updates, Telegram’s reputation for limited content oversight persists. Pairing the platform with Grok—an AI assistant already facing criticism for its loose filters—creates a dynamic that could quickly attract further scrutiny from regulators and watchdogs.
As Grok expands its presence, its performance, moderation, and monetization strategy will all be under a microscope.
