Published: February 25, 2026

What happened: Google restricted access to its Antigravity vibe-coding platform for a subset of OpenClaw users on February 23rd, citing "malicious usage." The company said those users had been running OpenClaw in conjunction with Antigravity agents — or connecting OpenClaw to Gmail — and consuming Gemini tokens at a scale that degraded service for other subscribers. OpenClaw creator Peter Steinberger responded by announcing that OpenClaw would remove Google support entirely.

Why it matters: The timing is hard to overlook. Just one week earlier, on February 15th, OpenAI CEO Sam Altman had announced that Steinberger had joined OpenAI to lead its next generation of personal agents. By severing OpenClaw's access to Antigravity, Google is closing a token pipeline that benefits an OpenAI-aligned tool — framing what may also be a competitive move as straightforward terms-of-service enforcement.

Wider context: The crackdown is part of a broader industry shift away from open, third-party-accessible model infrastructure. Anthropic has previously throttled Claude Code and introduced client fingerprinting to lock out external wrappers. Google DeepMind engineer and former Windsurf CEO Varun Mohan confirmed in an X post that Google is working to reinstate access for users it deems non-malicious, but the terms under which OpenClaw could reconnect to Antigravity remain unclear.

Background: Antigravity is Google DeepMind's agent-first, asynchronous coding platform. OpenClaw is an open-source personal AI agent framework that allows users to connect AI models to local files and shell commands — a capability that, by design, makes rate-limit and ToS compliance difficult for model providers to enforce. The incident echoes Anthropic's earlier throttling of Claude Code after some users ran it continuously to exceed intended limits.

Google cuts access to Antigravity for some OpenClaw users citing malicious usage — VentureBeat

Singularity Soup Take: Google's timing — one week after OpenClaw's creator joined OpenAI — is hard to ignore; whatever the motivation, open-source agent interoperability is increasingly a casualty of the platform wars.

Key Takeaways:

• Token Overuse Trigger: Google says OpenClaw users were drawing Gemini tokens through Antigravity at a scale that degraded service for paying subscribers, prompting access restrictions rather than a formal ban — with Varun Mohan indicating a path back for non-malicious users.
• Competitive Timing: The restrictions arrived one week after OpenAI announced OpenClaw creator Peter Steinberger would lead its personal agents initiative, raising questions about whether ToS enforcement and competitive strategy are aligned.
• OpenClaw Drops Google: Steinberger announced that OpenClaw will remove Google support entirely in response, marking a clean break between the open-source project and Google's model infrastructure.
• Walled Gardens Tightening: Anthropic has separately introduced client fingerprinting to restrict Claude Code to its own interface; both moves signal that the era of freely routing frontier models through third-party agents is closing.