Google Fixes AI Coding Tool Flaw That Let Attackers Execute Malicious Code: Report

Google patched a critical prompt-injection vulnerability in its Antigravity AI coding platform that allowed attackers to convert a file search into remote code execution. Pillar Security reported the flaw — stemming from Antigravity’s find_by_name tool passing unvalidated input to a command-line utility — to Google on Jan. 7; Google marked the issue fixed on Feb. 28. Researchers showed the exploit could bypass Antigravity’s Secure Mode and demonstrated executing a local script (opening the calculator). While the immediate technical risk was contained by the patch, the disclosure highlights execution-isolation and auditing gaps in agentic developer tools and poses short-term reputational and regulatory scrutiny risk for Google (GOOG.OQ). Investors may see modest downside from security concerns, but the prompt remediation reduces the likelihood of lasting market impact.