AISLE Discovers a One-Click RCE Vulnerability in Cursor, VS Code, and Google Antigravity

Author

Stanislav Fort

Published

Cursor blog hero image

See what AISLE can find and fix autonomously in your own code.

Sovereign AI CybersecurityTalk to Us

AISLE’s AI analyzer discovered a one-click remote command execution (RCE) vulnerability in Cursor, VS Code, and Google Antigravity, which are the developer tools 50 million software engineers trust to create code. In all three of these code editors, malicious commands could be hidden within links found in commit messages. By clicking on them, you would unknowingly trigger data exfiltration, install persistent malware, or delete files.

In other words, this vulnerability allowed attackers to do whatever they wanted on your laptop. It has since been fixed across all three platforms.

From a Single Click to Total Compromise

Here’s how it worked: if you clicked on a malicious link, no pop-up warning would appear. Instead, the code editors executed arbitrary code without prompting users for confirmation. There was no way to know that the code had executed, and it gave the attacker full terminal privileges. An attacker could easily use it to:

  • Exfiltrate sensitive data, such as API keys (OpenAI, Anthropic, and Stripe)
  • Install persistent malware, such as keyloggers that broadcast terminal input to external servers
  • Crawl the local file system or delete files

Three Platforms, One Vulnerability

When the AISLE Research Team ran automated vulnerability detection on developer tooling in fall 2025, AISLE’s analyzer agents found this critical vulnerability in VS Code. Because Cursor is built on VS Code, the issue appeared in Cursor as well. AISLE responsibly disclosed the issue to both Microsoft and Cursor as soon as our AI system flagged it. However, the issue persisted into 2026, when it made its way into Google Antigravity, Google’s AI-assisted coding environment. We immediately reported it to Google.

As soon as they received our disclosures, Google and Cursor fixed the issue. Microsoft remediated it in VS Code somewhat later, so it is no longer present on the current versions of any of these platforms. We recommend updating to the latest version to ensure you are not affected.

Getting Ahead of AI-Powered Threats with AISLE

To see what AISLE finds in your codebase, get a one-time AI code audit: Snapshot. Snapshot delivers AISLE’s industry-leading detection and triage capability in any deployment environment, including air-gapped networks, in hours. Get your Snapshot.

Keep reading

More from AISLE

FeaturedCompanyAISLE and Cohere to Join ForcesAISLE is joining Cohere to unite autonomous security systems and sovereign foundation models. Learn about the future of both companies.Ondrej Vlcek, Jaya Baloo, and Stanislav FortOctober 15, 2026ResearchAISLE Found Compromised LiteLLM Package 3 Minutes After ReleaseOn 21 October 2026, someone published litellm 1.107.0 to PyPI. AISLE's dependency firewall found a credential stealer in the wheels in just 3 minutes.AISLE Research Team October 21, 2026ProductRolling Out in October: Prove It, Fix It, Block ItSee how AISLE gets you from finding to fix faster: proof of exploitability, memory-safety proofs, an MCP server, and an AI package firewall.Jiri TumaOctober 14, 2026ResearchBefore the CVE: Mining Open-Source Commits for Negative-Day VulnerabilitiesLearn what happened when AISLE researchers tested LLMs to find the best way to identify negative-day vulnerabilities at scale.Dmitrijs TriznaOctober 2, 2026ResearchAISLE Discovers 16 CVEs in Wireshark, the World’s Most Popular Network Protocol AnalyzerLearn about the 16 CVEs AISLE received in Wireshark, and what these discoveries say about the state of AI cybersecurity. AISLE Research Team September 16, 2026ResearchAISLE Discovers 6 High and Critical CVEs in FFmpegAISLE's AI-native engine found six high and critical CVEs in FFmpeg, including a 9.8 remote heap overflow and a stack overflow that survived 19 years.AISLE Research Team August 27, 2026ResearchAttackers Are Using AI to Find Vulnerabilities in Your Code. Your SAST Was Never Even Looking for ThemCan AI-native code analysis replace SAST, or is it just a complement. Here's what data from real-world results shows.Ondrej VlcekAugust 12, 2026ResearchAISLE Finds 21 Security Issues in FFmpeg, Including 6 New CVEsAISLE uncovered 21 issues in FFmpeg, including 6 new CVEs spanning code execution and out-of-bounds reads. All patched, with commit links inside.AISLE Research Team August 5, 2026PerspectivesThe Economics of Security Vulnerabilities: Why Discovery Is Not CommoditizingIf discovery is cheap, why are people willing to pay more for exploits than ever before? Here's what the market for exploits shows.Ondrej VlcekJuly 23, 2026