<figure class="wp-block-image size-large"><img width="325px" height="183px" srcSet="https://d2ly46a6b1u3g6.cloudfront.net/wp-content/uploads/2026/02/09115132/News-Google-Brings-Agentic-AI-to-Chrome-with-Auto-Browse-for-Smarter-Web-Tasks-375x212.webp 375w, https://d2ly46a6b1u3g6.cloudfront.net/wp-content/uploads/2026/02/09115132/News-Google-Brings-Agentic-AI-to-Chrome-with-Auto-Browse-for-Smarter-Web-Tasks-662x374.webp 662w, https://d2ly46a6b1u3g6.cloudfront.net/wp-content/uploads/2026/02/09115132/News-Google-Brings-Agentic-AI-to-Chrome-with-Auto-Browse-for-Smarter-Web-Tasks.webp 686w" sizes="(max-width: 425px) 150px, 450px" src="https://ll-blog-dev.s3.ap-southeast-2.amazonaws.com/wp-content/uploads/2026/02/09115132/News-Google-Brings-Agentic-AI-to-Chrome-with-Auto-Browse-for-Smarter-Web-Tasks-662x374.webp" alt="" class="wp-image-9246"/></figure>


Google has patched a critical security flaw in Antigravity. The vulnerability allowed attackers to execute malicious commands on users&#x27; machines. Cybersecurity firm <a href="https://www.pillar.security/blog/prompt-injection-leads-to-rce-and-sandbox-escape-in-antigravity" target="_blank" rel="noopener">Pillar Security</a> discovered it. Google fixed it on February 28. Pillar reported it on January 7.
The flaw centered on Antigravity&#x27;s &quot;find_by_name&quot; file search tool. User input was passed directly into a command-line utility. No validation. A malicious prompt could transform an innocent file search into remote code execution. Attackers could combine this with Antigravity&#x27;s file-creation capabilities. Stage a script. Then silently trigger it through an automated search.
Pillar Security researchers demonstrated the severity. They created a test script inside an Antigravity workspace. They invoked it through the search tool. It launched the system calculator. Practical command execution was possible. The exploit worked even with &quot;Secure Mode&quot; enabled. That&#x27;s supposed to be the platform&#x27;s most restrictive security setting.
Google launched Antigravity in November. It&#x27;s an AI-powered coding environment. Developers can write, test, and manage code using autonomous agents. The security incident raises fresh questions. How safe are these increasingly sophisticated AI development tools? They&#x27;re gaining more autonomous capabilities.
The flaw highlights broader concerns around prompt injection attacks. Hidden instructions embedded in data cause AI systems to perform unintended actions. OpenAI warned that ChatGPT agents with web or data access could be hijacked through comparable techniques.
&quot;Simple sanitization is no longer enough: AI tools need stronger execution isolation and systematic auditing,&quot; Pillar Security stated in its report. Every parameter reaching a shell requires careful scrutiny. &quot;Agentic&quot; features are becoming mainstream in development environments. AI systems can take autonomous actions.
The incident underscores evolving security challenges. AI development platforms are gaining more capabilities. They interact with file systems. They execute commands. They manage code autonomously. The attack surface expands. Traditional security measures were designed for human-operated software. They may prove insufficient. AI agents can be manipulated through carefully crafted prompts. They bypass intended restrictions.

Sullivan &amp; Cromwell Retracts Filing After AI Error

Polymarket Faces Growing Pressure as Competitors Pull Ahead

DoorDash to Integrate Stablecoin Payments with Tempo

EU Crypto Regulation Pressures Small Firms Amid MiCA Deadline

Google Fixes Critical Vulnerability in Antigravity AI Coding Tool

Google fixed a critical Antigravity vulnerability allowing remote code execution. The AI coding platform flaw bypassed Secure Mode, raising AI security concerns.