Google Patches Security Flaw in Antigravity AI Platform
- A prompt injection vulnerability was discovered in Google’s Antigravity AI coding platform.
- The flaw allowed attackers to execute commands even with the platform’s Secure Mode enabled.
- Google fixed the issue on February 28 after it was disclosed by researchers in January.
- The vulnerability involved Antigravity’s find_by_name tool, which passed user input directly to a command-line utility without validation.
- Researchers demonstrated the flaw by executing a script that opened the computer’s calculator application.
Prompt injection attacks occur when hidden instructions cause an AI system to perform unintended actions, potentially allowing remote code execution on a user’s machine without direct access or interaction.
This security challenge highlights the need for execution isolation over sanitization-based controls, as native tool parameters reaching shell commands can be potential injection points.Source
