Google Limits Antigravity Access as Compute Costs Surge, Cuts Some OpenClaw Users’ Service

Google’s Antigravity platform, billed as a “vibe‑coding” environment for autonomous agents, has been throttled after a surge in what the company calls “malicious usage” that strained its Gemini token pool. According to a report by The Register, customers on the $250‑per‑month AI Ultra plan and lower‑tier subscribers have seen their accounts suspended without warning for running third‑party tools such as OpenClaw and OpenCode against Antigravity’s backend. The suspensions began in mid‑February and have affected a “subset of users” who were allegedly exceeding the intended usage limits of the service (The Register, Feb 23).

Google attributes the crackdown to a “massive increase in malicious usage of the Antigravity backend,” which it says degraded service quality for the broader user base. DeepMind engineer Varun Mohan, a former Windsurf founder, posted on X that the company needed a rapid way to cut off accounts “not using the product as intended” and that a remediation path would be offered for those who unintentionally violated the Terms of Service (The Register; VentureBeat). Mohan’s comments echo Google’s official narrative that the offending activity involved “accessing a larger number of Gemini tokens via third‑party platforms like OpenClaw,” a pattern that overwhelmed the system for other Antigravity customers (VentureBeat, Feb 23).

The timing of the enforcement move is noteworthy. Just a week earlier, OpenAI announced that OpenClaw creator Peter Steinberger had joined the company to lead its next generation of personal agents (VentureBeat). While OpenClaw remains an open‑source project under an independent foundation, its new financial backing by Google’s chief rival has heightened competitive sensitivities. By blocking OpenClaw‑derived traffic, Google is not only protecting its compute capacity but also limiting a conduit through which OpenAI‑adjacent tools could tap Google’s most advanced Gemini models. Industry observers have flagged the episode as a flashpoint in the broader tussle over autonomous‑agent ecosystems, where platform providers must balance open‑source innovation against the risk of uncontrolled token consumption (VentureBeat).

Developers caught in the sweep have taken to discussion forums to claim they were unaware of any contractual breach. The Register notes that many users “were unlikely to have actually read” the Terms of Service, yet the platform’s usage policies explicitly forbid “heavy autonomous usage” that is not provisioned or priced for the Antigravity service. Google’s enforcement, therefore, underscores a growing friction point: the need for clearer metering and billing models as AI agents become more self‑directed. Without such safeguards, providers risk sudden spikes in compute demand that can jeopardize service reliability for paying customers.

Google’s response also highlights an operational challenge for large‑scale AI infrastructure. Antigravity’s reliance on Gemini, Google’s flagship multimodal model, means that token‑level throttling can have outsized effects on downstream workloads. As Mohan explained, the company’s “limited capacity” forces it to prioritize “actual users” over experimental or hobbyist deployments that may inadvertently flood the system (The Register). The move may prompt other cloud AI vendors to tighten access controls on similar agent‑building platforms, especially as the market sees a proliferation of open‑source frameworks that can amplify token usage across multiple providers.

In the short term, affected developers can expect a “path for them to come back on” once they align with Google’s usage guidelines, according to Mohan’s X post. However, the episode serves as a cautionary tale for the nascent autonomous‑agent market: open‑source tools like OpenClaw can quickly become vectors for unanticipated compute loads, and platform owners are increasingly willing to enforce strict ToS compliance to protect their infrastructure and revenue streams.