Skip to content

Information Security News

  • Home
  • Cyber Security
  • Cybercriminals
  • Data Leak
  • Google
    • Android
  • Information Security
  • Linux
  • Malware
  • Microsoft
    • Windows
  • Open Source Tool
  • Vulnerability
  • Technology

Information Security News

  • Home
  • Cyber Security
  • Cybercriminals
  • Data Leak
  • Google
    • Android
  • Information Security
  • Linux
  • Malware
  • Microsoft
    • Windows
  • Open Source Tool
  • Vulnerability
  • Technology
  • Google

The Antigravity Crash: Google Purges $250/Month AI Subscribers for “Proxy Abuse”

by Nam Phong · February 25, 2026

Google has commenced a pervasive deactivation of accounts belonging to its premium subscribers, encompassing those invested in the AI Ultra tier at a monthly cost of $250. This disciplinary wave was precipitated by the utilization of company services via third-party conduits such as OpenClaw and OpenCode. These summary suspensions occurred devoid of prior admonition, sparking a surge of indignation among developers who are currently deliberating the fallout across various professional forums.

The crux of the controversy lies in a subset of users connecting third-party agentic interfaces to Google’s internal backend—specifically the Antigravity platform and Gemini CLI. Essentially, these subscriptions were transmuted into economical proxies for computationally intensive autonomous tasks, for which the service was never engineered, either technically or financially. Faced with a precipitous spike in overhead, Google opted for immediate, decisive intervention.

Varun Mohan, a co-founder of Windsurf and current DeepMind engineer, characterized these maneuvers as “malicious utilization,” asserting that they significantly degraded the quality of service for the broader user base. In a recent social media dispatch, he pledged to reinstate certain deactivated accounts—specifically those belonging to individuals who were ostensibly unaware of their transgressions. However, the affected community vehemently disputes this interpretation.

“Individuals paid for a quota, operated within the confines of those limits, and were subsequently banished. This does not constitute malicious use; it is the legitimate consumption of a product as marketed,” remarked AI engineer Mohan Prakash in a public post. He further noted that the service terms did not explicitly proscribe integration with OpenClaw, thereby leading users to reasonably assume its legitimacy. As a paradigm of corporate etiquette, he cited Anthropic, which, in analogous circumstances, simply issues an error message rather than resorting to total account termination.

This saga mirrors a recent episode involving Anthropic, which similarly penalized users for bridging subscriptions to external services in lieu of utilizing the more costly, official API. Both instances illuminate a systemic fragility within the sector: AI corporations are currently subsidizing tokens at rates significantly below their production cost to seize market share, gambling on future price escalations. When users engage with these services at high intensities, the underlying business model proves unsustainable. Ultimately, the cost of these structural failures is borne by the company’s reputation and the erosion of consumer trust.

Related coverage

  • Google Hit with Staggering Location Data Fine
  • Google Search Data Policy: AI Training and User Privacy
  • Google Sues Outsider Enterprise Over AI Phishing Scams
  • Google Faces Lawsuit Over Lyria 3 AI Music Training
  • Google ADK Vulnerability Enables Agent to Agent Attacks

Support Our Threat Intelligence

If you find our technology report and cybersecurity news helpful, consider supporting our work.

Buy Me a Coffee Logo Buy Me a Coffee PayPal
Crypto QR Code
USDT (TRC20):
TN8BdV8cp4T1Cd28gK9qTAnZknzzuwyUtm
USDT (ERC20):
0x3725e1a7d3bc5765499fa6aaafe307fabcd75bce

Tags: account suspensionAI tokensAI UltraAntigravitydeveloper news 2026GeminigoogleOpenClawTech ControversyVarun MohanWindsurf

Follow:

  • Next story RoguePilot: The Silent AI Hijacker Turning GitHub Issues into Repository Backdoors
  • Previous story Linus Torvalds Bumps Linux to Version 7.0 Because “Math is Hard”

  • Recent Posts
  • Popular Posts
  • Tags
  • Custom ChatGPT bot interface mimicking a service error to trigger the ClickFix RAT infection chain

    Malware

    ClickFix RAT Disguised as Custom ChatGPT Bot Strikes

    October 2, 2026

  • Google Threat Intelligence Group statistical dashboard showcasing AI-driven vulnerability discovery trends

    Vulnerability

    AI Discovery Shifts Focus to Critical Vulnerabilities

    October 2, 2026

  • Conceptual code snippet demonstrating the JSON Web Signature authentication bypass in Python Authlib

    Vulnerability

    Authlib Flaw Allows JSON Web Signature Authentication Bypass

    October 2, 2026

  • Conceptual representation of a teenager accessing the Microsoft Titan database infrastructure via an API vulnerability

    Vulnerability

    Teen Uncovers Flaw Exposing Trillions of Microsoft Records

    October 2, 2026

  • VUSec researchers demonstrating the Branch Target Reuse speculative execution attack on JIT compilers

    Vulnerability

    Spectre Variant BTR Exploits JIT Compilers

    October 2, 2026

  • Technique

    Power VPN Free VPN for Encrypted Browsing on a Windows Desktop

    September 24, 2026

  • OpenSUSE Leap 15.4 Beta releases, Linux distributions

    Linux

    OpenSUSE Leap 15.4 Beta releases, Linux distributions

    May 30, 2020

  • Ubuntu 16.04.6 LTS released: fix security vulnerabilities

    Linux

    Ubuntu 16.04.6 LTS released: fix security vulnerabilities

    March 1, 2019

  • GhostBSD 23.10.1 released, FreeBSD distribution

    Linux

    GhostBSD 23.10.1 released, FreeBSD distribution

    May 1, 2020

  • Solus 4.4 Fortitude releases, Linux distribution

    Linux

    Solus 4.4 Fortitude releases, Linux distribution

    January 26, 2020

  • AI AI security Android Apple APT BOTNET CISA cloud security Critical Infrastructure cryptocurrency cyberattack cybercrime Cyber Espionage cybersecurity Cybersecurity 2026 data breach DLL Sideloading Github google hacking Infosec InfoSec 2026 Infostealer Linux Linux Kernel malware Microsoft network security open source phishing privacy privilege escalation Prompt Injection ransomware RCE remote code execution security Social Engineering supply chain attack Tech News 2026 threat intelligence vulnerability windows Windows 11 zero-day
  • Home
  • About Us
  • Contact Us
  • DMCA NOTICE
  • Privacy Policy

Information Security News © 2026. All Rights Reserved.

Powered by  - Designed with Hueman Pro