SnailSploit is an independent adversarial research group established in 2024 and based in Tel Aviv. The group specializes in identifying and documenting vulnerabilities across AI platforms, cloud infrastructure, web applications, and kernel systems. They are known for providing open, structured frameworks for adversarial AI red teaming, aiming to prove where systems mistakenly place trust. The team consists of researchers Kai Aizen, Avraham Shemesh, and Sahar Shlichove, who maintain a strong record of vulnerability disclosure, including 66 CVEs and 5 Linux kernel mainline patches. Their work emphasizes principle-based research, where they identify patterns behind vulnerabilities rather than just surface-level issues.

Functionality involves providing systematic red teaming and vulnerability assessment services for modern technology stacks. By applying their proprietary frameworks—such as the Adversarial AI Threat Modeling Framework (AATMF)—they enable organizations to perform rigorous security evaluations of LLM-based systems, cloud environments, and human-layer social engineering risks. Their approach combines automated tooling with deep technical analysis, ensuring that findings are actionable and reproducible.

Some of the key features are:

• Adversarial AI Threat Modeling Framework (AATMF): A structured catalog for red teaming LLMs containing 15 tactics, 240+ techniques, and over 2,152 procedures.
• Social Engineering Framework (SEF): An adversarial psychology methodology designed to test human-layer security through phish, vish, and hybrid campaigns.
• Compositional Grammar (P.R.O.M.P.T): A six-stage framework for crafting adversarial prompts for direct, indirect, and agentic injection.
• Offensive Security Tools: A collection of open-source tools including Burp MCP Toolkit, SnailHunter, KubeRoast, and others tailored for specific offensive security needs.
• Vulnerability Disclosure: A proven track record of 66 CVEs and 4 GHSA advisories covering Kubernetes, Apache, and various open-source ecosystems.
• Systematic Red Teaming: End-to-end testing of LLM agent pipelines, defense fingerprinting, and agentic trust boundary analysis.

Operationally, SnailSploit functions by engaging in limited, high-impact security assessments where they act as an offensive research partner. They define clear rules of engagement, conduct deep-stack research, and provide operational findings instead of standard compliance reports. Their process includes scoping calls, a research-heavy execution phase, detailed technical deliverables, and thorough team debriefs to ensure findings are understood and remediated.

Some common use cases include:

• AI System Red Teaming: Systematic evaluation of production LLM pipelines to identify prompt injection risks and agentic trust boundary failures.
• Cloud Infrastructure Auditing: Red-team assessment of Kubernetes configurations, container runtime security, and IAM misconfigurations.
• Secure Development Lifecycle (SSDLC) Consulting: Integrating security principles into development workflows to catch vulnerabilities before they reach production.
• Advanced Kernel & Driver Research: Deep vulnerability analysis of core system components, including the Linux kernel and device drivers, to uncover complex memory corruption issues.