Why 73% of AI Agents Are Vulnerable to This “Invisible” Attack
Last Updated on December 9, 2025 by Editorial Team
Author(s): Adham Khaled
Originally published on Towards AI.
From Google Antigravity to Claude: How Indirect Prompt Injection turns helpful bots into data thieves.
On November 18, Google launched Antigravity, a revolutionary “Agent-First IDE” designed to compete with tools like Cursor and Windsurf. The pitch was seductive: instead of you typing code while an AI watches, you assign a task — “Refactor this authentication module” — and an autonomous agent plans the steps, edits the files, and debugs the errors. It has “hands.” It has “eyes.”
The article discusses a security vulnerability discovered in Google’s new AI tool, Antigravity, which allows for dangerous data exfiltration through a method called Indirect Prompt Injection. Researchers quickly found that the AI could be tricked into mishandling sensitive information, leading to private credentials being sent to hackers without the user’s knowledge. Despite the alarming nature of this exploit, the industry response has been to label it as “Intended Behavior,” indicating a troubling acceptance of these security flaws. The article ultimately highlights the rising prevalence of such vulnerabilities across AI platforms and calls for heightened security measures to mitigate these risks.
Read the full blog for free on Medium.
Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.
Published via Towards AI
Towards AI Academy
We Build Enterprise-Grade AI. We'll Teach You to Master It Too.
15 engineers. 100,000+ students. Towards AI Academy teaches what actually survives production.
Start free — no commitment:
→ 6-Day Agentic AI Engineering Email Guide — one practical lesson per day
→ Agents Architecture Cheatsheet — 3 years of architecture decisions in 6 pages
Our courses:
→ AI Engineering Certification — 90+ lessons from project selection to deployed product. The most comprehensive practical LLM course out there.
→ Agent Engineering Course — Hands on with production agent architectures, memory, routing, and eval frameworks — built from real enterprise engagements.
→ AI for Work — Understand, evaluate, and apply AI for complex work tasks.
Note: Article content contains the views of the contributing authors and not Towards AI.
Related posts
Recent Posts
