Technology intelligence for a changing world

About · Editorial standards

CyberProfi

ENGLISH EDITION

AI powers discovery of Ghostscript ‘Upload2Shell’ vulnerability

AI discovers Ghostscript vulnerability: Researcher uses AI to identify critical Upload2Shell flaw in 38-year-old software, raising urgent cybersecurity.

In a breakthrough for cybersecurity and artificial intelligence, a security researcher has used advanced AI methods to discover a critical vulnerability—dubbed Upload2Shell—in the widely deployed Ghostscript software, first released 38 years ago. The finding underscores both the growing capabilities of AI for good and the heightened risks as attackers can wield similar tools. This news was first reported in several authoritative outlets on November 22, 2026, with primary confirmation from the RIO Rundown, KPTV, and AI industry news sites.

This article explores how artificial intelligence led to the identification of the Upload2Shell vulnerability, its technical implications, responsible disclosure, and what this event signals for software security in an age where AI is rapidly changing the threat landscape.

AI discovers Ghostscript vulnerability: what happened?

On November 22, 2026, cybersecurity researcher Paulos Yibelo announced (per RIO Rundown) the discovery of a severe exploit in Ghostscript—a print and document conversion toolkit used across web servers, publishing systems, Linux distributions, and enterprise platforms. Yibelo utilized artificial intelligence to probe complex legacy code, successfully identifying an Upload2Shell flaw allowing remote code execution by uploading specially crafted font files. Ghostscript’s long history and ubiquity make this a high-risk, high-impact vulnerability for organizations worldwide.

The research leveraged deep learning to map control flow and data paths, flagging sequences that previously evaded conventional static analysis. This marked one of the first publicly documented cases where AI independently discovered an exploitable zero-day in a mature, open-source project.

Technical background: What is Upload2Shell?

The core of the Upload2Shell vulnerability lies within Ghostscript’s font-parsing modules. By uploading a manipulated font (OTF or TTF), attackers could trigger a bug resulting in arbitrary command execution on the system running the Ghostscript process. Systems processing user-submitted documents or images—such as web-to-PDF services, publishing platforms, and cloud conversion APIs—are especially at risk.

Security advisories (see RIO Rundown and Artificial Intelligence News) note that exploitation is possible wherever Ghostscript’s sandbox fails. Yibelo’s AI model mapped known and hidden code paths, scanning for edge cases traditional fuzzers missed. This multiplies both the ability of defenders to anticipate flaws and, crucially, the potential for sophisticated attackers to automate similar exploits.

Response and industry impact

The researcher coordinated disclosure to Ghostscript’s maintainers, who issued an emergency patch across supported branches and notified major Linux and cloud vendors. U.S. authorities cited the incident in emphasizing the growing interplay of AI with cybersecurity—both for rapid exploitation and for protection (see KPTV coverage). In addition, the event triggered broad software updates at many hosting providers, including those offering embedded PDF conversion services and online form processing.

Technical leaders and CISOs now face more direct questions: How will AI-driven scanning redefine responsible disclosure? Should defenders assume previously trusted code bases may still harbor machine-findable flaws? Will attackers now regularly deploy similar AI tools at scale, tipping the balance between vulnerability disclosure and exploitation risk?

How AI found what humans missed

Ghostscript has been under active development since the late 1980s and subject to frequent audits by both open-source contributors and major security vendors. Yet, as reported by multiple sources, no prior scanner or audit tool had uncovered the Upload2Shell flaw. Yibelo’s use of generative AI and advanced code embeddings allowed the model to reason through non-linear code execution paths and identify dangerous memory patterns.

This approach signals a new era, with generative models performing static and dynamic analysis at a scale and depth previously reserved for large security teams—now accessible to independent researchers or, potentially, well-resourced adversaries.

Risk assessment and recommendations

  • Patch immediately: Organizations running Ghostscript should update to the latest patched release. Web-to-PDF, online form, and publishing services are top priority.
  • Audit supply chain: Review indirect dependencies. Ghostscript is embedded in many content management, document handling, and automation systems; update all affected packages.
  • Monitor for anomalous uploads: Platforms enabling document or image uploads must scan logs for suspicious font-related activity.
  • Review privilege configurations: Run Ghostscript as a non-privileged user and inside containers where possible to reduce attack surface.
  • Adopt AI-driven code review: Consider modern AI tools that augment code security audits and threat detection, applying lessons learned from this incident.

For further insight, review CyberProfi’s archive of cybersecurity and artificial intelligence category articles that outline best practices for responding to zero-days and using AI securely.

Broader implications: AI and the security arms race

The Upload2Shell case arrives as U.S. government officials, including members of the Senate’s Cybersecurity Caucus, highlight the dual-use nature of AI in security. Increasingly, AI enables both defenders and attackers to operate at a new scale, rapidly scanning and weaponizing old and new code alike.

According to recent press coverage, Google’s own cybersecurity reports stress the sophistication of modern AI-driven attacks, including those that now automate steps once requiring painstaking manual work (KPTV). Responsible research, as shown in the Ghostscript event, must keep pace with the potential for rapid exploitation.

Looking ahead, organizations must assume that AI-powered security testing will become a routine requirement and that AI-based attack tools will proliferate. Cybersecurity teams should update incident response plans, invest in ongoing code review, and work closely with vendors to secure the entire supply chain.

Frequently asked questions (FAQ)

What is Ghostscript and why is the Upload2Shell vulnerability critical?
Ghostscript is a widely-used interpreter for PDF and PostScript files, integral to print, scan, and conversion workflows. The Upload2Shell vulnerability lets attackers execute code on affected systems, posing serious risks for any platform handling uploaded documents or images.
How did AI identify the vulnerability?
The researcher used AI-driven static analysis to model Ghostscript’s code flow and spot conditions for exploit, going beyond human review and prior automated tools.
Is a patch available?
Yes. Ghostscript maintainers released updates immediately following responsible disclosure. Users should patch all vulnerable installations and dependencies.
Does this mean AI will find more vulnerabilities?
Almost certainly. As AI matures, both defenders and attackers can expect the rate of discovery to increase, especially across legacy or complex code bases.

Sources