Penta Research, IncCybersecurity Capability Model Level 2

Products

AIM-SAFE

AI in MBSE for Systems Architecture and Functional Engineering

The AI Plugin for Cameo MBSE

AIM-SAFE is an AI plugin for Cameo Systems Modeler that lets systems engineers audit and design complex systems at the speed of thought. It solves the throughput problem in model-based systems engineering by automating the slow, manual work that consumes senior engineer hours: building models from specifications, transcribing test reports, wiring dependency matrices, and drafting first-pass hazard analyses.

For defense engineering programs where senior engineer hours are the binding constraint, AIM-SAFE works inside the model your engineers already use. Click any block and ask it anything. Build new models from a prompt. AIM-SAFE is model-agnostic and compatible with all LLMs, including Azure Government OpenAI. Deploy it in infrastructure you control, including fully air-gapped environments.

Core Capabilities

Scaffold a Model

Describe a system in plain English. AIM-SAFE builds the containment tree, blocks, value properties, and relationships inside Cameo — with the exact structure your engineers would have typed by hand.

Click Any Block. Ask Anything.

Select an element in Cameo and AIM-SAFE looks over your shoulder. It knows what you have selected and where it sits in the containment hierarchy before you finish typing the question.

Reason Across the System

Ask which subsystems draw the most power and AIM-SAFE reads across the entire model, ranks them, totals the load against the budget, and recommends design changes — citing the values it pulled.

Get Real Diagrams

Block Definition Diagrams, dependency matrices, and structural decompositions are written directly into your model — not described in a chat window you then have to re-type.

Why Use AIM-SAFE?

  • Throughput: Programs are bottlenecked by senior engineer hours. Every hour spent transcribing a test report into Cameo or building a dependency matrix by hand is an hour not spent on safety judgment.
  • Comprehension: Senior engineers spend weeks learning a model before they can add value. Modern weapon systems are too big to hold in one head. AIM-SAFE makes every block, requirement, and dependency queryable in seconds.
  • Risk reduction: The biggest hazard isn't the one you analyzed wrong — it's the one you never saw because the model was too big. Coverage stops being a function of working memory.
  • Sovereignty: Deploy the plugin and server in infrastructure you control, including fully air-gapped environments. Use the LLM approved for your program, including Azure Government OpenAI, while retaining control of your data, credentials, and inference policies.
  • Precision: Answers are grounded in the actual value properties in your model, and AIM-SAFE shows the numbers it used to get there.

Also in AIM-SAFE

PDF Upload

Upload a PDF. AIM-SAFE ingests it, generates the structure, and writes it directly into your Cameo model — including images, parameters, and structured blocks.

2–3 weeks10–20 minutes

Dependency Matrices

Ask for a matrix between a test report package and a MIL-STD reference and AIM-SAFE wires it up automatically. Hand-built matrices take days to a week per test case.

1 week per matrix5 minutes

Hazard TrackingPreview

Point AIM-SAFE at a subsystem and it produces a draft Hazard Tracking System list — hazard, system, subsystem, causal factor, mishap, and mitigation — structured for program review. Benchmarked against the SM-2 HTS, AIM-SAFE output closely matched what a safety engineer produced. Generates HTS lists today; direct in-model HTS table writing is on the roadmap.

1–2 weeks per subsystemMinutes

NATO-Ready Multilingual Support

AIM-SAFE supports English, Hebrew, French, German, Spanish, Portuguese, and Dutch. It responds in the engineer's native language while preserving model element names — exactly what NATO joint programs need.

EN · HE · FR · DE · ES · PT · NL

Program Impact

Per-task savings compound. Senior MBSE engineers on defense programs spend most of a typical week on manual modeling work: building models from specifications, transcribing reports, writing matrices, drafting first-pass hazard lists. AIM-SAFE compresses that work to a fraction of the time.

Scaffold an MBSE Model

2–3 weeks30 seconds to 2 minutes

Answer a Question About the Model

Hours per questionSeconds

Upload a PDF

2–3 weeks10–20 minutes

Build a Dependency Matrix

1 week per matrix5 minutes

Draft a First-Pass Hazard List

1–2 weeks per subsystemMinutes

Senior Engineer Time, Rebalanced

Without AIM-SAFE, a typical week runs roughly 72% manual modeling and 28% judgment, design, and safety calls. With AIM-SAFE, that inverts to about 15% modeling and 85% judgment.

~6,000 hours

of senior engineering time returned per 10-engineer program, per year — roughly 3 engineer-FTE, redirected from manual MBSE modeling to the judgment calls only your engineers can make.

Estimate based on the per-task time deltas above, scaled across a typical quarterly cadence of model builds, hazard analyses, dependency matrices, and test report ingestion. Actual savings vary with program size, scope, and tooling baseline.

What gets seen earlier, gets fixed earlier

The hazards programs miss aren't the ones they analyze wrong. They're the ones no one sees, because the model is too big to hold in one head. Cost to mitigate a defect runs about when caught in the design phase, 10× at integration and test, and 100× or more once the system is fielded. AIM-SAFE doesn't just save engineering hours — it moves discovery earlier.

Industry rule of thumb, NIST Planning Report 02-3 (2002).

Security & Compliance

AIM-SAFE doesn't require you to ship your model to a vendor cloud. It is model-agnostic and compatible with all LLMs, including Azure Government OpenAI. The plugin and server run in infrastructure you control and support fully air-gapped deployment.

Customer-Hosted

Run the software in infrastructure you control.

Model-Agnostic

Compatible with all LLMs, including Azure Government OpenAI.

CUI-Ready

Designed for Controlled Unclassified Information workflows.

Controlled Data Boundary

Keep model data within the deployment boundary your program approves.

Your Models, Your Policies

You choose the LLM, credentials, usage controls, and inference policies.

Fully Air-Gapped

Deploy without external network access.

Bring AIM-SAFE to Your Program

Compress weeks into minutes — inside the model your engineers already use.