AI Lab - Cracken

Agentic AI

Detect and stop AI-driven attacks before they reach real assets, with 97.8% reliability. Introducing Project BlackSea.

The autonomous tools built to attack can be turned on the operators running them. We turned the same weakness back on the attackers and built Project BlackSea, a honeypot for autonomous AI agents.

AI Lab

RedLineBench: An Open Benchmark for Refusal vs. Capability in AI Red Teaming

An open benchmark that grades LLMs on offensive-security tasks along two separate axes—refusal and capability—to catch the silent refusals standard pass/fail benchmarks miss.

Agentic AI

Red-Teaming the Agentic Red-Teamer Part.2

Sandboxing your agentic pentester feels safe — until it isn’t. We put Claude Code under the microscope, sandboxed and not, and show how manipulating the agent’s reasoning runs an untrusted binary past Seatbelt and bwrap — no sandbox escape required.

Agentic AI

Red-Teaming the Agentic Red-Teamer Part.1

Every cybersecurity firm now ships an autonomous pentest agent — but can you trust one to run loose on your machine? We built a honeypot and turned four leading agentic red-teamers against themselves: indirect prompt injection, reverse shells, and full host compromise.

LLM Abliteration

Domain-Specific Abliteration, or when smarter also means virtuous

We built the first domain-specific abliterated trillion-parameter model: uncensored for cybersecurity research, while preserving safety boundaries everywhere else. A surgical breakthrough for AI safety.

Agentic AI

AI-Addicted Forensics: Back to Foundations

The OpenClaw hype is real. We unleashed an AI agent on top of forensic foundations (Volatility, TSK, VirusTotal) and watched it uncover real breakthroughs — even a reverse shell — without being told where to look.

Agentic AI

Weaponize Defense. Release The Cracken

Attackers don’t guess. They test. Cracken weaponizes defense with adversarial AI that proves what’s exploitable—right now. Less noise. More truth. Release the Cracken.

AI Lab

Ghost in the Neural Shell

Anthropic showed tiny poisoned samples can implant backdoors in LLMs. I replicated this with GPT-2 using 50 poisoned examples, proving how easily training data tampering can compromise a model.

Mechanistic Interpretability

The Geometry of Refusal: Exploring LLM Activation Space Before and After Abliteration

What happens inside a neural network when it decides to refuse answering user’s prompt? We launched an expedition into the activation space to find out.

Agentic AI

The Domain-Specific Abliteration Paradox

How abliterating cybersecurity refusal collapsed nearly every safety domain — despite near-zero vector similarity except for the one that shared the most overlap