Summary
code injection in gem (CVE-2026-41512). Successful exploitation can lead to full system takeover. Exploitable via POST /targets/auto_detect_selectors.
Original description
The description as published by the source.
ai-scanner is an AI model safety scanner built on NVIDIA garak. From version 1.0.0 to before version 1.4.1, there is a remote code execution vulnerability via JavaScript injection in BrowserAutomation::PlaywrightService. This issue has been patched in version 1.4.1.
Source: NIST National Vulnerability Database NVD
AI summary
Written by an external AI model and may contain mistakes. Check the original description and references below before acting on it.
A critical vulnerability has been identified in the AI model safety tool "ai-scanner". This issue allows attackers to inject malicious code, resulting in arbitrary execution without requiring user intervention. The vulnerability affects versions from 1.0.0 to before 1.4.1, and upgrading to version 1.4.1 resolves the issue. Immediate update is strongly advised if your company uses this tool.
Between versions 1.0.0 and 1.4.1 of ai-scanner, a RCE vulnerability exists via JavaScript injection in BrowserAutomation::PlaywrightService. The gateway is the POST /targets/auto_detect_selectors endpoint where URLs are improperly escaped before insertion into Node.js scripts, allowing execution of arbitrary code. The issue is patched in 1.4.1. Update command: gem update ai-scanner. No workaround has been provided, immediate update is required.
A remote code execution vulnerability through JavaScript injection.
In BrowserAutomation::PlaywrightService via POST /targets/auto_detect_selectors, specifically by interpolating unescaped URLs.
Critical risk due to potential remote code execution without user intervention.
Update to ai-scanner version 1.4.1.
No workaround specified.
Monitor for unusual requests to POST /targets/auto_detect_selectors with suspicious URLs.
Related past incidents Similar incidents extracted from past CVEs
Similar RCE via unescaped input leading to arbitrary code execution in web applications.
Pulse Connect Secure Remote Code Execution Vulnerability (similar execution path exploitation).
Although different in nature, Heartbleed is an infamous example of a critical exploit.
If this happens at your company Expected impact per business scenario
Potential unauthorized access to critical systems and sensitive information exposure due to RCE vulnerability.
Exploitation can lead to lateral movement across Docker containers, compromising multiple services.
Service disruption and potential data breach leading to loss of trust and financial penalties.
Organizations should update their ai-scanner installations to version 1.4.1 immediately and review security settings.
Response actions
Concrete steps and command examples for SOC/SRE teams to execute in order
-
1Identify exposure identify
Audit SBOM/dependencies for affected components.依存マニフェストで影響コンポーネントを特定する。
-
2Match against affected range verify
Confirm if version satisfies `>= 1.0.0, <= 1.4.0`Step 1 で見つかったバージョンが影響範囲
>= 1.0.0, <= 1.4.0に該当するか照合。本番で稼働中ならインシデント扱い。 -
3Hunt for indicators of compromise detect
grep 'POST /targets/auto_detect_selectors' /var/log/nginx/access.log | grep -E '(unusual_payload|sqli_pattern)'アクセスログで
POST /targets/auto_detect_selectorsへの異常なリクエスト (不正な認証ヘッダ・SQLメタ文字)を過去 30〜90日分捜索。WAF/SIEM があれば該当パスのアラート発火履歴を確認。 -
7Post-deployment verification verify
Replay attack against POST /targets/auto_detect_selectors on staging to confirm patch closes the vectorパッチ適用後、ステージングで PoC または同等の悪用パターンを再現して脆弱性が閉じたことを確認。本番では Step 3 と同じログクエリでアラート再発が無いか継続監視。
References
- exploit github.com
- patch security-advisories@github.com