Google Bans $250/Month AI Subscribers for Connecting OpenClaw - No Warning, No Refund
Google disabled Antigravity accounts for paying subscribers who linked OpenClaw via OAuth, citing service degradation. Creator Peter Steinberger calls it draconian.

Google is banning its own paying customers for using an open-source tool. AI Ultra subscribers paying $250 per month for Antigravity access are getting their accounts disabled without warning, without a grace period, and without a refund - because they connected OpenClaw via OAuth.
The ban wave started around February 12-14 and has been escalating since. Affected users see a single message: "This service has been disabled in this account for violation of Terms of Service." No email. No explanation. Just a 403 PERMISSION_DENIED error where their $250/month service used to be.
TL;DR
| Detail | Value |
|---|---|
| Who's affected | Google AI Pro and Ultra subscribers ($250/mo) |
| Trigger | Connecting OpenClaw via Google's OAuth tokens |
| Error | 403 PERMISSION_DENIED, "violation of Terms of Service" |
| Warning given | None |
| Refund offered | None reported |
| Ban wave started | ~February 12-14, 2026 |
| Google's reason | "Massive increase in malicious usage" degrading service |
| OpenClaw response | Creator will remove Antigravity support |
What happened
OpenClaw, the open-source AI agent framework with 219,000+ GitHub stars, lets users connect their own AI provider accounts via OAuth. When users linked their Google Antigravity subscriptions, OpenClaw routed requests through Google's private OAuth client ID - making the traffic appear as if it came from an official Google product rather than a third-party application.
The result was a flood of API requests that Google's systems flagged as abuse. Varun Mohan, the lead for Google Antigravity, confirmed the crackdown:
"We've been seeing a massive increase in malicious usage of the Antigravity backend that has tremendously degraded the quality of service."
The issue is real. OpenClaw agents generate far more API calls than a human user typing in a chat interface - spiraling tool calls, recursive reasoning loops, and multi-step task execution can burn through what a normal user consumes in a day within minutes. When thousands of OpenClaw users hit the same subsidized consumer endpoint simultaneously, the infrastructure buckles.
How the ban works
Google's detection method appears to rely on anomalous usage patterns: request volumes, call frequencies, and tool-use patterns that don't match typical IDE or chat usage. Once flagged, the account is disabled immediately. No graduated response. No throttling first. No "first and final warning."
A Google employee on Hacker News clarified that the bans affect only Antigravity and Gemini CLI access - not Gmail, Drive, or other Google services. But for users who subscribed specifically for AI access, that distinction is cold comfort.
The support nightmare
What's making this worse than a standard ToS enforcement is the support experience. Users report:
- No pre-ban warning - no email, no notification, no chance to stop using OpenClaw first
- No refund - subscribers continue to be charged for a disabled service
- No clear reinstatement path - support told one $250/month customer "we are unable to reverse the suspension" after 11 days of trying
Mohan acknowledged there should be "a path for them to come back on" for users who were unaware of the ToS implications, but added the caveat of "limited capacity" - suggesting reinstatements are being handled manually rather than systematically.
Google Support told a $250/month Ultra subscriber: "We are unable to reverse the suspension." Eleven days. No resolution.
Steinberger's response
Peter Steinberger, OpenClaw's creator (who joined OpenAI on February 14), called the ban wave "pretty draconian" and said he plans to remove Antigravity support from OpenClaw entirely. His argument: users connected their own paid accounts through a standard OAuth flow, and Google should throttle rather than ban.
The timing is notable. Steinberger joined OpenAI the same week the bans started. OpenAI, which acquired influence over the project, has already whitelisted OpenClaw's traffic through its own API.
A pattern across providers
Google isn't alone. Two of the three largest AI providers locked down third-party access in the same week:
| Provider | Action | Date |
|---|---|---|
| Anthropic | Updated legal terms to explicitly ban OAuth token usage in third-party tools; implemented client fingerprinting | Late January 2026 |
| Disabled Antigravity accounts for OpenClaw-linked users without warning | Feb 12-14, 2026 | |
| OpenAI | Whitelisted OpenClaw after acquiring creator | Feb 14, 2026 |
The fundamental tension is economic. Consumer AI subscriptions are heavily subsidized - Google, Anthropic, and OpenAI all lose money on their $20-$250/month plans, betting on user acquisition. When an AI agent tool turns one consumer subscription into an automated API pipeline, it turns the unit economics from "subsidized user acquisition" to "burning cash on programmatic access that should be billed at API rates."
The real question
The Hacker News discussion crystalized the core debate. One side argues Google has every right to enforce its ToS: users knowingly delegated their credentials to a tool that violated usage terms, and the resulting traffic degraded service for legitimate users. As one commenter put it: "They are too stupid to understand the implications."
The other side points out that banning paying customers without warning - while continuing to charge them - is exactly the kind of behavior that makes developers afraid to build on Google's platforms:
"I won't subscribe to Google despite good models. The enforcement approach damages customer relationships regardless of technical merit."
The pragmatic middle ground: throttle first, warn second, ban last. A "first and final warning" approach would have preserved the same outcome - getting OpenClaw traffic off the subsidized endpoint - without the PR damage and the support backlog of angry customers who paid $250 for a service they can no longer use.
OpenClaw's growing list of collateral damage now includes not just security incidents but platform bans affecting thousands of paying users. For an open-source project with 219,000 GitHub stars and a creator who just joined OpenAI, the question isn't whether OpenClaw is powerful enough to disrupt AI providers' business models. It's whether the providers will respond with rate limits or with bans.
Sources:
- Google Antigravity ban wave hits OpenClaw users, creator weighs ending support - PiunikaWeb
- Google restricting Antigravity users for using OpenClaw - Hacker News
- Google banning accounts for using OpenClaw with Antigravity - GitHub Issue #14203
- $250/mo Ultra Subscriber Banned Without Warning - Google AI Developers Forum
- Google restricts Antigravity users tied to suspicious OpenClaw behaviour - BusinessToday
- Google Restricts AI Ultra Subscribers Over OpenClaw OAuth - Implicator
- Account restricted without warning - Google AI Developers Forum
