OpenClaw, the once-unthinkable leap in generative intelligence, signals a moment when AI stops imitating creativity and starts competing with it.
When it was first out, called Clawdbot, it was ignored. When it renamed itself Moltbot, it was shining but still underestimated. Now, as OpenClaw, it no longer has that problem. Its advancements reflect how quickly generative AI can evolve from prototype to pressure point. And because of that, it has entered the conversation with a sharper edge and far bigger implications.
OpenClaw, unlike others, is able to reshape how stories are told, images are made, and entire industries are challenged in real time.
And Google doesn't like that.
The tech titan took a decisive action against a surge of what it described as malicious usage on its Antigravity AI coding platform, resulting in restricted access for numerous users who had integrated the open-source AI agent OpenClaw.
This move, which unfolded over the weekend, stemmed from concerns over system overload and degraded service quality for legitimate subscribers.
Read: Moltbot Molts Again And Becomes 'OpenClaw,' And How Cloudflare Steps In With Moltworker
We’ve been seeing a massive increase in malicious usage of the Anitgravity backend that has tremendously degraded the quality of service for our users. We needed to find a path to quickly shut off access to these users that are not using the product as intended. We understand…
— Varun Mohan (@_mohansolo) February 23, 2026
Google DeepMind engineer Varun Mohan explained the situation clearly in a public statement: a massive increase in such exploitative patterns had overwhelmed compute resources intended for paying subscribers, violating the platform's terms of service.
While not every user intended harm, some were likely unaware that third-party agent integrations in this manner contravened the intended use. Google prioritized protecting overall service integrity and fairness and because of that, the company emphasized that restrictions targeted Antigravity access specifically, not broader Google services like Gmail, though some users reported wider disruptions possibly tied to shared account systems.
Google has indicated plans to reinstate certain unaware users under controlled conditions, acknowledging limited capacity to handle appeals promptly.
A final note - we deeply regret not giving more warning for this but we needed to move fast to preserve quality for all our users and took action on accounts where a vast majority of usage (>90%) was not the Antigravity product. We will do better next time and are working hard to…
— Varun Mohan (@_mohansolo) February 24, 2026
Antigravity serves as Google's innovative "vibe coding" environment, leveraging its Gemini AI models to assist with development tasks through an agent-first architecture. It caters especially to paying users on plans like the $250-per-month AI Ultra tier, who expect reliable performance and fair resource allocation.
OpenClaw, meanwhile, is a popular open-source autonomous AI agent framework that runs locally or on personal servers.
It empowers users to create powerful personal assistants capable of executing complex workflows, calling tools, running shell commands, and managing files. Its appeal lies in true agentic behavior, where the AI can autonomously chain actions and interact with external services, often far more intensively than standard chat interfaces.
The conflict arose when developers began connecting their paid Gemini accounts to OpenClaw via OAuth credentials, using plugins like the one for Antigravity.
This setup allowed OpenClaw to route requests through Google's backend, effectively turning the flat-rate subscription into a high-volume proxy for the agent.
Unlike per-token API billing designed for heavy or automated usage, these consumer-oriented plans are priced and provisioned for more moderate, interactive sessions. OpenClaw's autonomous nature, capable of burning through millions of tokens in short bursts or tens of thousands on seemingly simple queries, amplified consumption dramatically.
Automated loops, multi-step reasoning chains, and persistent background operations pushed servers to their limits, flooding the Antigravity backend and causing noticeable degradation for other users relying on the same shared infrastructure.
Pretty draconian from Google. Be careful out there if you use Antigravity. I guess I'll remove support.
Even Anthropic pings me and is nice about issues. Google just... bans? https://t.co/JBq9YCB7nB— Peter Steinberger (@steipete) February 23, 2026
This isn't an isolated incident; it echoes Anthropic's recent tightening of rules around Claude integrations, where similar OAuth token routing for third-party agents led to throttling or bans.
The pattern reflects a broader industry shift as frontier AI providers grapple with the economics of flat-rate subscriptions versus the explosive demands of autonomous agents. Viral tools like OpenClaw highlight the tension between open-source innovation and platform sustainability.
Developers love the flexibility and power, but unchecked high-volume access can strain subsidized infrastructure built for lighter, human-paced interactions.
In response, OpenClaw's creator, Peter Steinberger, criticized Google's approach as overly harsh and announced plans to drop Antigravity OAuth support to avoid further issues for the community. The episode underscores risks in depending on third-party wrappers for frontier models: without explicit approval or metered billing, such integrations can lead to sudden enforcement actions.