Measured intelligence
MizanIQ is not a chat assistant — it is the legal intelligence engine that MizanLabs products run on, with a single rule: no claim without a source, and no action without human approval. Every intelligent capability in our products states its limits before its promises.
How it thinks
The chain of proof
Every result carries the thread of its source, from the official page to your screen — from claim to evidence in one click.
Confidence in layers, not percentages
We do not show percentages of false precision. Confidence is shown as described levels, and whatever is ambiguous goes to human review.
No training on your data
Your firm’s and your clients’ data never enters the training of any model — learning means rules and retrieval inside your firm alone.
A declared autonomy ladder
Every capability states its level: detects, explains, proposes, prepares — and execution is always in human hands.
Questions people actually ask
Does it give me legal advice?
No. MizanIQ detects, extracts and summarises what has been published officially, and shows the source of every line. It does not issue opinions and does not derive a legal rule of its own — any legal information appearing in our products comes with its text and its reference, or it does not appear.
Is it a language model trained on Egyptian law?
With us, the law is not stored in a model’s weights. The text changes and has an effective date, an edition and a jurisdiction, so it is retrieved at answering time with its text and its source. Models perform the task; they do not confer legal authority.
Why don’t you publish an accuracy percentage?
Because a single percentage applied to an individual fact promises a precision we cannot measure for that fact. We show instead the confidence layer, described, together with the source of the result, so the lawyer judges for themselves — that judgement is what we build for, not what we replace.
What happens when it does not know?
It says “I don’t know.” An absence is shown as an absence, and whatever is ambiguous is routed to human review rather than filled in with a guess that looks confident.
Security first?
The architecture of isolation, auditing and the limits of the intelligence is set out in full on the trust page.