NEW

Start with the pressure: sales, launch, abuse, agents, data, or guardrails

Workshops

AI Security Workshopsthat leave artifacts, not slide decks.

Productized working sessions for product security, AI platform, red team, blue team, governance, and executive teams. Use them to plan the program, review the architecture, run adversarial tests, build detection, or clean up public claims.

Workshop planner

Five scoped sessions, one operating surface.

Productized

Program planning

AI Security Program Planning Jumpstart

3 hours

Teams that need a first AI security program, not a generic strategy deck.

program charterstakeholder map

Architecture review

AI Security Architecture Review Sprint

4 hours

RAG, agent, and model-driven features that need a concrete review before launch.

threat modelcontrol backlog

Red team

AI Red Team Jumpstart

1 day

Teams that need their first adversarial test plan and evidence pack.

attack planfindings memo

Blue team

AI Blue Team Jumpstart

1 day

Teams that need visibility before they can tune response.

event schemadetection backlog

Governance

Governance and Claim-Readiness Jumpstart

3 hours

Teams that need trust language without inflated maturity language.

claim registerevidence matrix

Artifacts you leave with

threat model
control backlog
architecture memo

5 formats

Program planning, architecture, red team, blue team, governance

3 hours to 1 day

Sized for real teams and real calendars

Public-safe outputs

Backlogs, memos, playbooks, evidence maps

Cross-functional

CISO, AppSec, platform, SOC, GRC, and product

Artifacts

Backlogs, memos, playbooks, evidence maps

Workshop formats

Pick the session that matches the problem.

Each workshop is scoped to a real decision point: starting the program, reviewing a feature, testing abuse paths, building detection, or making claims defensible.

Architecture review

AI Security Architecture Review Sprint

Threat-model one AI feature or product path with data flows, trust boundaries, control gaps, and evidence requirements.

4 hours
Design review
AppSec leaders, AI platform engineers, security architects, product teams

Best for

RAG, agent, and model-driven features that need a concrete review before launch.

Outcomes

threat modelcontrol mapfix list

Session agenda

1

System map

Inventory inputs, outputs, models, tools, retrieval paths, and trust boundaries.

45 min

Output: A single shared view of the system.

2

Attack paths

Walk through likely abuse cases, data leaks, unsafe actions, and authorization failures.

55 min

Output: A ranked list of abuse paths.

3

Control design

Map controls to the risks that matter: logging, approvals, eval gates, and guardrails.

50 min

Output: Controls tied to actual failure modes.

4

Review memo

Package the findings, owners, and next engineering moves for follow-up.

30 min

Output: A short memo that is ready to circulate.

threat modelcontrol backlogarchitecture memoreview notes

Public-safe outputs

Every session ends with something your team can use.

Backlog

Prioritized work with owners, urgency, and next moves.

Memo

Concise decision artifact for executives and engineering.

Evidence map

Claims connected to controls, records, and artifacts.

Playbook

Repeatable steps for detection, response, or review.

Control map

Risks connected to controls, gaps, and evidence requirements.

Retest plan

Minimum follow-up checks and checkpoints after fixes.

Delivery flow

Structured like a planner. Delivered like a working session.

Every workshop starts with a clear scope, moves through concrete decisions, and ends with an artifact that can be handed to engineering, leadership, or operations.

01

Choose the workshop

Pick the format that matches the problem: planning, architecture, red team, blue team, or governance.

02

Scope the system

Name the target, the people in the room, and the evidence you want to leave with.

03

Run the session

Move through the agenda live, with a bias toward decisions instead of slides.

04

Package the output

Leave with a backlog, memo, evidence map, or playbook that can move into execution.

Next step

Pick the workshop that matches the problem, then scope it once.

The point is to produce useful artifacts, not to sell a vague training day.