OpenAI opens ChatGPT subscriptions to OpenClaw’s 3.2M users as Anthropic blocks Claude’s access to the AI ​​agent platform.

Sam Altman posted on X at 2:33 am on May 2: “You can login to openclaw with your chatgpt account now and use your subscription there! Lobsterings are fun.” The announcement, delivered with the usual register of an inventor pushing a small product update, is not so trivial. OpenAI has made its ChatGPT subscription the authentication and payment layer of OpenClaw, an open AI agent framework that has become the fastest growing project in the history of GitHub, with 346,000 collected in less than five million months and in less than five months it is now used by more than five million people. ChatGPT subscribers Plus they can log in with OAuth, access GPT-5.4 with a Codex endpoint, and run autonomous AI agents on their own hardware for $23 a month in total OpenAI hasn’t built the world’s most popular AI agent.

Lobster

OpenClaw was founded in November 2025 by Peter Steinberger, an Austrian developer who once sold a software company for $100 million and was experimenting with AI coding tools in a Madrid cafe. The first version was called Clawdbot, a play on Anthropic’s Claude with a lobster mascot. Anthropic filed a trademark complaint. Steinberger renamed it Moltbot, and then, because that “didn’t roll off the tongue,” he renamed it OpenClaw. The lobster stayed.

The product is a locally hosted AI agent that connects to major language models, Claude, GPT, DeepSeek, and others, and works with messaging apps that people already use: WhatsApp, Telegram, Signal, Discord, Slack, iMessage, Microsoft Groups. It manages calendars, sends emails, organizes files, writes code, browses the web, and automates multi-step workflows. The data resides on the user’s machine. The agent runs continuously in the background. Jensen Huang called it “is the most popular open source project in human history” at Nvidia’s GTC conference in March It surpassed GitHub’s decade-old record for React by 60 days.

In February, Altman announced that Steinberger was joining OpenAI to “drive the next generation of personal agents” and that OpenClaw would be moved to an independent foundation with continued support and funding from OpenAI. Sequoia distributed 200 Mac Minis at the AI ​​event as OpenClaw became the infrastructure layer for potential capitalists, and the signal from the most influential firms in Silicon Valley was clear: the layer of agents would open, and business models would have to be built around it rather than on top of it.

An opposite bet

On April 4, Anthropic blocked Claude Pro and Max subscribers from using their low-cost subscription plans with OpenClaw and other third-party AI frameworks. The reason was cost: OpenClaw’s automated agents can generate thousands of API calls per day, using far more computing power than a human typing questions into a chat window. Anthropic decided that access to unlimited subscriptions through an agent framework was not economically sustainable and was closing.

Anthropic’s decision to block OpenClaw from Claude’s registration was a precautionary measure to protect margins. OpenAI’s decision to do the opposite, to open the ChatGPT subscription to OpenClaw, is annoying. By making ChatGPT the default backup for the world’s most popular agent framework, OpenAI is betting that the volume of new subscribers will more than offset the increase in computing costs per user. The economics only work if OpenClaw converts a significant number of its 3.2 million users to paying ChatGPT subscribers. If it does, OpenAI will have found a distribution channel for its subscription product for which no marketing value has been created.

Competitive strength is prominent. Anthropic looked at OpenClaw and saw a cost problem. OpenAI looked at a similar product and saw an opportunity to distribute it. Another company closed the door. He opened the other one and took out the keys.

Accidents

OpenClaw’s rapid growth has been accompanied by equally rapid security failures. At the end of January, a critical remote code execution vulnerability, CVE-2026-25253, was disclosed: any website visited by a user can silently connect to a local proxy server via an unauthenticated WebSocket, which includes cross-site hijacking for full code execution on the user’s machine. Security researchers checked ClawHub, OpenClaw’s skills marketplace, and found 824 verified malicious entries out of 10,700 available skills, 335 of which were traced to a single coordinated attack. More than 30,000 instances of OpenClaw found exposed on the public Internet without verification. Moltbook, a social layer for agents, suffered a breach that exposed 1.5 million API tokens and thousands of private conversations.

The vulnerability has been patched in current versions. The problem is that a significant portion of the installed base uses old, unpatched versions. Anything before version 2026.1.30 remains vulnerable to at least some of the vulnerabilities, and attackers still target it. OpenAI’s decision to merge its ChatGPT subscription into OpenClaw means that the OpenAI brand, its payment system, and its user data now flow through an open source platform that has had more security incidents in four months than most enterprise software in a decade.

The ecosystem

Nvidia has transformed OpenClaw into an enterprise platform with NemoClaw, adding enhanced security, compliance features, and integration with Nvidia’s indexing infrastructure. Tencent launched ClawPro, an enterprise AI agent platform built on the OpenClaw architecture and designed for the Chinese market. Meta launched Manus AI as a desktop agent, a competitive approach that works like a native app rather than messaging apps. The agent layer is now a battleground where every major technology company is taking a stand.

ChatGPT’s subscription integration puts OpenAI at the center of this ecosystem without requiring it to own or manage the agent framework itself. OpenClaw remains open source, governed by an independent foundation, and compatible with multiple language model providers. But with Anthropic’s access to blocking and OpenAI enabling it, the practical result is that three million OpenClaw users are switched to ChatGPT as their default model. The foundation structure gives OpenAI deniability. Subscription integration gives you distribution.

Model

Economics is not normal. A ChatGPT Plus subscription costs $20 per month. OpenClaw Launch Lite, a hosted management layer, costs $3 per month. For $23, the user gets access to GPT-5.4 through the OpenClaw agent framework without the API cost per token. This is much cheaper than using the OpenAI API directly, which can cost hundreds of dollars per month for the volume generated by an independent agent. OpenAI supports agent usage through its subscription tier, betting that the lifetime value of a subscriber using ChatGPT with OpenClaw is higher than the combined cost of serving their agent requests.

This is the same idea that drove mobile carriers to subsidize smartphones: to offer the hardware economy to offset subscription revenue. OpenAI gives the agent access to lock the ChatGPT subscription. If the bet works, ChatGPT becomes not just a chatbot but an automated intelligence layer for a generation of AI agents that control people’s digital lives. If it doesn’t work, OpenAI will be opening up its most valuable product to a computationally intensive use case that’s fueled by speculative volume without generating any measurable revenue.

Altman’s tweet was seven words and a lobster joke. The decision behind it is one of the most consistent distribution bets OpenAI has made since launching ChatGPT. The most popular open source project in history is now running on your ChatGPT subscription. Whether that’s a masterstroke or a margin trap depends entirely on whether the three million lobster enthusiasts convert to paying customers, and whether the agent they use on their laptops is secure enough to be trusted by OpenAI and its subscribers.