Meet Selma.

The AI that never leaves the building.

Get in touch Watch the overview

Meet Selma

Selma is AI that runs entirely inside your organisation. On your hardware, trained on your knowledge, founded on your values, and never connected to the internet.

Our belief

The organisations doing the most important work are the ones most locked out of AI.

The reason is simple: every major AI tool sends your data to overseas servers. For organisations handling sensitive client records, legally protected information, or confidential research, that's not a trade-off they can make.

That's why we built Selma. Founded on an organisation's values, installed in their building, trained on their knowledge, and never connected to the internet.

It's AI that never leaves the building.

There's a safer way to use AI.
It's called Selma.

Selma is the AI for organisations unwilling to compromise.

Why

AI can transform the way organisations work. Most can't use it.

Not because they don't want to. Because the tools available demand compromises they're not willing to make: data sent to overseas servers, energy costs that conflict with sustainability commitments, outputs that can't be fully trusted, and models that absorb your most sensitive knowledge with no guarantee of where it goes. For the organisations doing the most important work, the cost has always been too high.

Selma removes the compromises. Not the capability.

What

An alternative AI, built from your values up.

Selma is an alternative AI solution. One that begins with your values, not your data. We understand your beliefs, identify the right use cases, and install a custom instance on your premises, fine-tuned to your knowledge and principles. Every contribution is attributed, and every response reflects who you are.

Data sovereignty is the floor.
Selma is the ceiling.

How

Built for your building.

EthicAI installs the hardware on your premises, works with your team to embed your ethical constitution in the model, and sets up attribution tracking for every staff contribution. Your team accesses everything through a simple browser on your internal network.

No data leaves the building.
Every response reflects your values.
You remain in control.

Pre-seed round · Moving to close Q3 2026

Selma is being built right now.

If you want to be part of it, register your interest below.

Register interest

Inside Selma

Ethical constitution
Your values and principles, embedded directly in the model. Selma reasons within your ethical framework.
Knowledge attribution
Every staff contribution tracked, attributed, and rewarded. The model remembers who did the teaching. (In development)
Air-gapped architecture
No internet connection. No data transfer. Ever. Entirely within your local network.
On-premises installation
Hardware lives in your building. EthicAI installs and hands it over. What happens next is entirely yours.
Fine-tuned to your knowledge
Trained on your documents. Selma knows your policies, your language, and the way your organisation thinks.
LAN-accessible
Your team accesses Selma through any browser on your network. No app, no external login, no internet required.
Scroll to explore

Organisations Selma is designed for

01

Ombudsman and complaints bodies

Organisations receiving highly personal disclosures and bound by strict investigation protocols. Selma drafts correspondence, surfaces relevant policy, and supports case management without a single byte of complainant data leaving the building.

02

Community legal centres

Legal aid services operating under strict professional privilege and privacy obligations. Selma assists with document drafting, research, and client intake, grounded entirely in the organisation's own legal knowledge base.

03

High-value IP organisations

Research firms, advisory practices, and specialist consultancies whose entire value is what they know. Feeding that knowledge into a public model isn't a risk to be managed — it's a strategic loss. Selma trains on your knowledge base and keeps it entirely yours.

04

Child protection agencies

Agencies handling some of the most legally protected information in the public sector. Selma supports caseworkers with document preparation and policy lookup in an environment where data sovereignty is non-negotiable.

05

Policy advisory bodies

Organisations advising government on sensitive matters including national security, budget, and social policy. The knowledge they produce is too valuable and too sensitive to feed into an external model. Selma keeps it sovereign.

06

Professional regulators

Regulatory bodies managing confidential disciplinary and complaint processes. Selma supports investigation and decision-making workflows while keeping the identity of all parties entirely within the organisation's walls.

Founded on your values.
Installed in your building.
Never connected to the internet.

Who We Are

EthicAI.

6M+
TikTok views in under a year
AI Summit
Global AI Impact Summit, Australian Government delegation
GIA
Selected by the Governance Institute of Australia to develop national AI board training
Live
Hardware platform operational and ready for deployment

EthicAI is a Sydney-based AI ethics and governance consultancy built on a simple conviction: AI adoption should never compromise an organisation's values, its legal obligations, or its control of its own data. We joined the Australian Government delegation to the Global AI Impact Summit in Delhi, presented at the Parliamentary Innovation Showcase in Canberra, and have operationalised MIT Sloan's EPOCH human capability framework. We have also built a genuine public audience, passing over six million TikTok views in under a year. Few teams sit credibly at both ends of this conversation, the policy table and the public feed.

Selma is what that expertise has become. We work inside the environments Selma is built for, which means we understand the constraints from the ground up rather than consulting on them from a distance.

Kara Bombell
Kara Bombell
Co-founder
Katriel Healy
Katriel Healy
Co-founder
Simon Brock
Simon Brock
Co-founder

"The next great populist movement... it won't be anti-immigration or anti-globalisation. It'll be anti-AI... This is an enormous opportunity."

Scott Galloway, Professor of Marketing, NYU Stern

FAQs

Is Selma a real product or a concept?

Selma is a working system. The hardware platform is operational and the software is in active development. The pre-seed round funds completion of the product and our go-to-market infrastructure.

What are the details of the round?

We are running a pre-seed round, moving to close Q3 2026. Register your interest below and we'll be in touch with a deck and full round details.

Who are the target customers?

Selma is built for organisations that handle sensitive, legally protected, or commercially valuable information and cannot use cloud-based AI tools. Initial targets include ombudsman bodies, community legal centres, child protection agencies, policy advisory bodies, and high-value IP organisations across Australia.

What makes Selma defensible?

The defensibility is not in the model, it's in the expertise behind the deployment. EthicAI has spent years building ethics and governance frameworks, helping organisations across sectors navigate AI adoption responsibly and within local legal principles. That domain knowledge is embedded in how Selma is configured, customised, and governed for each client. We also have a local footprint and deep relationships with the kinds of organisations Selma is built for. A larger player could build the hardware. They cannot replicate years of on-the-ground experience helping complex organisations do this well.

What will the investment be used for?

Four things: completing the product, developing sales and marketing infrastructure, pursuing IP protection for the Knowledge Attribution Model (KAM), and securing the talent needed to scale delivery. The timeline: product complete Q3 2026, go-to-market Q4 2026, IP filing in parallel.

What is the Knowledge Attribution Model (KAM)?

When staff use and refine Selma, they are teaching it. The Knowledge Attribution Model (KAM) tracks exactly what each individual contributes to the model's growing capability, weighted by impact on performance. This gives organisations a mechanism to understand, and reward, who is actually building their AI. No equivalent system exists in the industry, and we are pursuing IP protection for it.

Why now?

The conversation around AI sovereignty and the potential of small language models has been building for some time. Organisations across the public sector and regulated industries have been grappling with data obligations that make cloud AI untenable, and the infrastructure argument for running AI on local hardware has grown steadily stronger. But the urgency became impossible to ignore in June 2026, when the US Commerce Department ordered Anthropic to suspend access to its Claude Fable model overnight. Every customer globally lost access with no warning. Australian organisations had no recourse. That event made concrete what had previously felt like a theoretical risk: when your AI infrastructure runs on someone else's servers in another country, the switch is not yours to control. The shift to sovereign, on-premises AI is not a future trend. It's a present need.

Why is it called Selma?

Selma is named in honour of Selma Lagerlöf, the Swedish author who became the first woman to win the Nobel Prize in Literature, in 1909. Lagerlöf proved that language, used with intention and values at its core, can create genuine and lasting change in the world. That conviction sits at the heart of what we're building. It happens that Selma also stands for Small Language Model (SLM), the architecture at the centre of the product.