Consulting
Consulting with security in the room.
Most AI programmes stall in a security review nobody can pass, or a pilot nobody can operate. We work both sides, with the same team.
What we usually find
The problem is rarely the model.
- Shadow AI nobody tracked
- Prompt injection and data leaks
- No compliance evidence
- Too many AI agents, no one watching
What we do
Assess, design, harden, prove.
- Finding your AI usage
- Security testing for AI systems
- A secure blueprint to build from
- Safety checks and testing
- Compliance evidence
- Locking down the systems around it
What you get
Documents that survive contact with reality.
A list of what's running, a security review, a working reference build, a set of tests, and an evidence pack mapped to your controls.
Who you are working with
Engineers first.
29Eleven Technologies is a small consultancy of senior engineers who built and defended systems at FAANG scale. Working on platforms with hundreds of millions of users teaches you which complexity earns its keep, so that is where your budget goes: identity, data boundaries, evaluation, cost control and the runbook for the bad day.
Senior only
The person who scopes the work does the work. We would rather turn down a project than staff it with people who need to learn on your systems.
We write code
Strategy decks are a by-product. The deliverable is a system running in your environment, with tests, dashboards and a runbook.
Security is not a phase
Threat modelling happens while the architecture is still cheap to change, not in a review three days before launch.
We hand things over
Every engagement ends with your team able to operate what we built. Retainers should be a choice, not a dependency.
We say no
If a model is the wrong tool, or the data is not ready, or a cheaper fix exists, you will hear it early and in writing.
Plain language
You should be able to read our documents to your board and your auditor without a translation layer.
Start here
Start with an assessment.
A short, scoped assessment tends to settle arguments that have been running for months about what is safe to ship. Tell us what you are working with and the first call can begin with your systems instead of introductions.