Google RCE flaw patched in AI antigravity tool

The recent discovery of a Google RCE flaw in its AI‑driven antigravity utility exposed a prompt‑injection bug that let attackers break out of the sandbox and execute arbitrary code.

Google RCE flaw patched after urgent security review

The vulnerability stemmed from insufficient sanitization of user prompts, creating a remote code execution vector. Security researchers from Reuters disclosed the issue last week, prompting a rapid response. Google’s engineering team rolled out a fix that tightens input validation and adds runtime checks.

“The patch closes the vector that allowed malicious commands to be injected,” a Google spokesperson said.

Analysts note that the ↓ 1 critical vulnerability has been neutralized, but the episode underscores the expanding attack surface of generative AI tools. Ongoing vigilance remains essential as AI integrates deeper into system utilities.