Google has restricted access to its Antigravity “vibe coding” platform for some users, citing “malicious usage” and a surge in activity linked to the open-source AI agent OpenClaw. The move, which began over the weekend and continued Monday, February 23rd, resulted in some users losing access to their Google accounts, according to reports on social media.
According to Google, the restrictions stem from users leveraging Antigravity to access a disproportionately large number of Gemini tokens through platforms like OpenClaw, overwhelming the system for other customers. Varun Mohan, a Google DeepMind engineer and former CEO of Windsurf, explained in a post on X that the company observed a “massive increase in malicious usage” that degraded service quality. “We needed to find a path to quickly shut off access to these users that are not using the product as intended,” Mohan wrote, adding that Google is working to reinstate access for users unaware of the policy violations.
The timing of Google’s action is notable, coming just one week after OpenAI announced that Peter Steinberger, the creator of OpenClaw, had joined the company to lead its “next generation of personal agents.” While OpenClaw remains an open-source project managed by an independent foundation, its strategic direction is now influenced by Google’s primary competitor. By limiting OpenClaw’s access to Antigravity, Google is not only addressing server load concerns but also disrupting a pathway for OpenAI to utilize Google’s Gemini models, according to industry observers.
The restrictions have impacted a range of users, including those paying $250 per month for Google’s AI Ultra subscription, as well as those with less expensive plans. Some users reported losing access to their Google accounts entirely after running OpenClaw instances connected to Google products, as discussed on Y Combinator chat boards and X. Google DeepMind has clarified that access has only been restricted to Antigravity, not other Google applications.
Steinberger responded to the restrictions by announcing that OpenClaw will remove Google support. The incident highlights the challenges of interoperability and trust when integrating open-source AI agents with proprietary platforms. Google’s move is not intended as a permanent ban on third-party platform access, but rather an effort to align usage with its terms of service, a Google DeepMind spokesperson told VentureBeat.
This is not an isolated incident. Last year, Anthropic throttled access to Claude Code after identifying similar abusive usage patterns. The current situation underscores a broader industry trend toward “walled garden” agent ecosystems, where providers prioritize vertically integrated experiences and control over telemetry and revenue. Anthropic recently introduced “client fingerprinting” to ensure Claude Code remains the exclusive interface for its models, effectively blocking third-party wrappers like OpenClaw.
For enterprise technical decision-makers, the “Antigravity Ban” illustrates the risks of relying on agentic dependencies. The incident demonstrates the fragility of platform access, even for high-paying customers, and the potential for third-party wrappers to create vulnerabilities. Experts suggest prioritizing agent frameworks that can run “local-first” or within Virtual Private Clouds (VPCs) and decoupling AI development from core corporate identity providers to mitigate risks associated with Terms of Service violations.
