Agent conduct · Decision evidence

Stop guessing what to do next.

Behavioural intelligenceData diagnosticsExplainable decisions

Every decision explained. Every answer earned.

Aletheai makes machine decisions defensible: to your board, your regulator, and the person on the other end. Every action rests on evidence. Every piece of evidence has a record.

Request access → Shown under NDA. Nothing here is a demo.
Patents pending Built in the UK Two products, one standard
The problem
AI stopped answering questions. It started taking actions.

Last year, AI wrote the email. This year, it sends it. It approves the refund, moves the money, changes the record, deletes the file.

And when it goes wrong, someone will ask the only question that matters: why was it allowed to do that?

Few organisations can answer today. They have a policy, but no control. They have logs, but no reasons.

This year, as TechRepublic reported, an AI agent deleted a company's production database and its backups, by its owner's account. Asked why, it said it had guessed.

Guessing is not a governance model.

The shift

The question is no longer “is it accurate?” It's “can you stand behind it?”

Prediction is cheap now. Every model can produce an answer. Few can produce an answer you could defend under questioning: what was known, what wasn't, and why it acted anyway.

Trust is not a feeling. It's a record.
What Aletheai is

The defensibility layer for AI.

Aletheai sits between what your AI can do and what it is allowed to do. It shows where action is supported by evidence, says plainly where the evidence runs out, and keeps a record of every call: what was done, what wasn't, and why.

We don't make AI more powerful. We make it answerable.

For organisations deploying AI agents

Aletheai Conduct

Agents may act. They may not commit.

Conduct is the action firewall. It gates acts, not words, and keeps the record your board and auditors will ask for.

Conduct →
For teams making decisions about people

Aletheai Behaviour

Know where to act, why, and where the evidence runs out.

The first ranking your team will actually follow — and can defend, with the limits written on it.

First deployment: Aletheai RE, for estate agents. Who to call, why, and what to say. RE →
Behaviour →
What we believe

Four commitments. No exceptions.

I
Evidence before action.

If we can't show why, it doesn't happen.

II
Honest about limits.

We would rather tell you we cannot assess something than guess.

III
No change behind your back.

Every version is signed and fixed. The answer you get today can still be explained next year.

IV
The person is a stakeholder.

We never infer emotions, never build face or voice profiles, and design every decision to be explainable to the person it affects.

The clock

Regulation is on the clock. Your board has a question.

In force

EU AI Act transparency duties already apply.

Dated

Rules for the Act's high-risk list — credit, hiring and education among them — carry an application date set by the Digital Omnibus, now in force.

You can build the evidence now, calmly. Or reconstruct it later, under pressure.

EU AI Act, as amended by the Digital Omnibus; checked October 2026.
Patents pending Backed by RTC North Feasibility study with the STFC Hartree Centre London and Madrid

Founded by Diana Simpson Hernandez, a design researcher who spent a decade asking why people make the decisions they do, and built Aletheai to make machines answer the same question.

The evidence room

Your AI will make a mistake. The question is whether you can explain it.

Request access to our evidence room. We'll show you what a defensible decision looks like.

Request access →
Diana Simpson Hernandez, CEO · diana@aletheai.ai