We do not build systems that watch people

There is a large market for software that observes people and reports what it saw. We are not in it, and this page exists so that is a matter of record rather than a matter of trust.

Our systems work from what a person gives them directly. They do not collect behavioural signal in the background in order to form a view about someone, and they do not exist to tell a third party what a person did.

If our software cannot tell something by asking, it does not know it, and it says so rather than inferring.

Observation changes the person observed. A system that watches produces a subject who performs for it, which makes the observation worse evidence than the conversation it replaced. The technique is self-defeating before it is anything else.

It also produces the wrong relationship. Software that reports on you is not working for you, whatever the interface says, and people are correct to treat it accordingly.

• Collecting activity signal for the purpose of characterising a person.
• Producing suspicion, risk scores, or flags about individuals.
• Any capability whose primary value is telling an institution what one of its people did.