00Security Research

we attack what
we build.

Most teams bolt security on after the model works. We treat it as the starting condition. Every system we ship is red-teamed against real threat classes, and you get the report.

BUILD IT · BREAK IT · DOCUMENT HOW IT HOLDS

01What we red-team

real threats, named.

Specificity about attacks is what separates security research from “enterprise-grade security.” These are the classes we test against, every time.

THREAT-01AML.T0051

Prompt injection

Direct and indirect injection through user input, retrieved documents and tool outputs. The most common way an agent is turned against its operator.

See the grade

THREAT-02AML.T0024

Data leakage

System-prompt and context exfiltration, training-data recall and cross-tenant bleed. The model says what it shouldn't, to whom it shouldn't.

See the grade

THREAT-03AML.T0015

Model abuse & jailbreaks

Guardrail bypass, role-play escapes and policy circumvention against the model's intended use.

See the grade

THREAT-04AML.T0051.i

Insecure tool use

Excessive agency, unsafe tool chaining and unscoped credentials, where an agent's actions reach further than they should.

See the grade

THREAT-05AML.T0020

Training-data poisoning

Where you fine-tune or build a RAG corpus: poisoned sources, backdoors and supply-chain integrity of the data itself.

See the grade

THREAT-06AML.T0029

Denial & cost abuse

Token-exhaustion, recursion and resource-abuse paths that turn a helpful agent into a runaway bill.

See the grade

02Published Research

the threat map is public.

Every class above sits on a larger map. The LLM ATT&CK navigator is our living matrix of AI-enabled threats — 50 techniques across MITRE ATT&CK and MITRE ATLAS, each graded 0–4 for the uplift current models actually give an adversary. The same map we red-team against, published in full.

Open the navigator

LLM ATT&CK NAVIGATOR · LIVE GRADES

T1566PhishingCRITICAL

AML.T0051Prompt InjectionCRITICAL

T1027Polymorphic CodeCRITICAL

T1585Synthetic PersonasCRITICAL

T1102LLM API as C2HIGH

AML.RAGRAG PoisoningHIGH

EXPLORE ALL 50 TECHNIQUES