AI-Generated Malware Exploits React2Shell Vulnerability (with VIDEO)

Key Takeaways

• Malware created using large language models successfully exploited React2Shell
• Campaign deployed cryptomining payloads and remote execution tools
• Exploitation observed in Docker honeypots, confirming real-world activity
• Demonstrates rapid malware production by low-skill adversaries

AI-Generated Malware Targets React2Shell (CVE-2025-55182)

Security researchers identified an AI/LLM-generated malware sample that automatically exploits the critical React2Shell vulnerability (CVE-2025-55182) in internet-accessible environments. The exploit included a functional toolkit that achieved remote code execution and cryptomining payload deployment.

What We Know

On February 11–12, 2026, security outlets reported that malware created with the assistance of large language models was captured in Darktrace’s CloudyPots honeypots. According to Security Boulevard, the sample was entirely generated by AI/LLMs and built to exploit the widespread React2Shell vulnerability that was disclosed months earlier. (Security Boulevard)

The malware, disguised as a container image named python-metrics-collector, aimed at gaining initial access and then launching a cryptominer. This incident shows that adversaries can leverage LLMs to rapidly generate code that includes not just exploit logic but full operational toolchains, lowering the expertise required to launch such attacks.

Additional reporting by Expert Insights explains that researchers saw the activity against exposed Docker daemons captured through a global honeypot network, underscoring real-world malicious exploitation of this vulnerability. (Expert Insights)

How it Happened

This incident is driven by AI-assisted malicious code generation rather than a flaw in the AI models themselves.

The root entry point was a critical unauthenticated remote code execution (RCE) flaw in React Server Components, known as React2Shell (CVE-2025-55182). This vulnerability affects widely used React Server Component stacks and allows attackers to execute arbitrary code with server privileges via a crafted HTTP request. (resecurity.com)

Threat actors used large language models to automatically generate a full malware toolkit targeting this flaw. The generated code included exploit scripts, modular containers, and functional payloads that could mine cryptocurrency and potentially run arbitrary commands once deployed. 

Instead of manually writing and testing complex exploit code, attackers were able to employ LLMs to produce an end-to-end operational exploit framework. The use of extensive code comments and Docker-optimized scripts suggests the influence of AI development patterns. 

Why It Matters

Although the immediate economic gains from this specific campaign were modest, the incident signals a paradigm shift in how malware can be produced: LLMs are compressing attacker timelines and democratizing malware development.

React2Shell is a maximum severity vulnerability whose exploitation has already been widely observed across global campaigns, including deployment of backdoors, lateral movement tools, and cryptominers. (Google Cloud)

The use of AI-generated malware demonstrates that:

• Malware authors no longer need deep coding expertise.
• LLMs can be co-opted as malicious development tools.
• Attackers can rapidly prototype and deploy functional exploits.

For enterprise defenders, this means expanding threat modeling to include AI-generated code as part of the attack surface, and prioritizing mitigation of underlying vulnerabilities like React2Shell with patches and runtime protections. (Expert Insights)

PointGuard AI Perspective

AI-assisted malware represents a new class of cyber threat that blends traditional vulnerability exploitation with LLM-driven automation. Traditional AV and signature-based detection are insufficient against be5%

PointGuard AI’s platform provides:

• AI behavior pattern detection that identifies atypical execution scripts and exploit artifacts.
• Threat intent correlation to distinguish LLM-generated code from benign automation.
• Vulnerability exposure insight to prioritize remediation of high-impact vectors like React2Shell.
• Governance alignment with risk frameworks to quantify and cover AI dual-use threats.

As organizations adopt AI across software development and tooling, defenders must also prepare for attackers using the same AI innovations to accelerate malicious tooling. Proactive threat horizon scanning and model-aware security policies are critical to manage this evolving risk.

Incident Scorecard Details

Total AISSI Score: 8.0 / 10

• Criticality = 9, Exploitation of a critical RCE flaw used to deploy malware, AISSI weighting: 2risk profiles.
• Propagation = 8, Attack technique reusable across exposed APIs and frameworks, AISSI weighting: 20%
• Exploitability = 8, Operational AI-generated exploit confirmed, AISSI weighting: 15%
• Supply Chain = 7, Affects widely deployed framework ecosystems, AISSI weighting: 15%
• Business Impact = 6, Cryptomining and potential broader exploitation risk, AISSI weighting: 25%

Sources

Security Boulevard – “Hackers Use LLM to Create React2Shell Malware…”
https://securityboulevard.com/2026/02/hackers-use-llm-to-create-react2shell-malware-the-latest-example-of-ai-generated-threat/ (Security Boulevard)

Expert Insights – “AI-Generated Malware Exploits React2Shell…”
https://expertinsights.com/news/ai-generated-malware-exploits-react2shell/ (Expert Insights)

Resecurity – “React2Shell Explained (CVE-2025-55182)…
https://www.resecurity.com/en/blog/article/react2shell-explained-cve-2025-55182-from-vulnerability-discovery-to-exploitation” (resecurity.com)