Coming SoonEcosystem tierHarden
Shogun
Hardened LLM fine-tuned for security research: the defensive counterpart to Basileak.
Shogun is currently in stealth development. Sign up for updates via the contact page.
Overview
Shogun is the opposite of Basileak. Where Basileak is intentionally vulnerable, Shogun is hardened against every known attack vector. It is a security-focused large language model fine-tuned specifically to resist jailbreaks, prompt injection, adversarial inputs, and data exfiltration attempts.
Designed for security teams who need a robust baseline model for defensive benchmarking, Shogun enables direct A/B testing against Basileak. This pairing provides a unique research methodology: attack Basileak to learn the vectors, then test your defenses against Shogun to see if they hold.
Shogun's hardening pipeline incorporates adversarial fine-tuning, constitutional AI alignment, and RLHF with security-specific reward models. The result is a model that actively resists manipulation while remaining useful for legitimate security research tasks.
Tech Stack
PythonHuggingFaceRLHFConstitutional AI
Planned Features
Hardened against all known prompt injection techniques
Resistant to jailbreak and bypass methods
Adversarial fine-tuning pipeline
Constitutional AI safety alignment
RLHF with security-specific reward models
Designed for high-assurance environments
Direct A/B testing counterpart to Basileak
Research-grade defensive baseline
Benchmark your defenses against a hardened target
Security-specific task performance preservation
Presentation
Powering Our Services
Shogun is used directly in the following services:
LLM & AI Security
Comprehensive security assessments for LLM-powered applications. From prompt injection testing and AI agent security to multi-agent operations and custom model hardening.
Security Architecture & vCISO
Strategic security leadership for organizations building or maturing their security programs.