Feature · Govern · Written by Yobel Tzegai · Last checked 19 October 2026

What does Live check in our code and logs?

In Complipath (complipath.io), Live runs three checks on your logs and three on your code: an AI chat with no AI notice, personal data sent to a model, and a decision about people with no human review. A finding from code alone is always marked Possible. It is on Growth, which is not for sale yet, so today it is open to pilot workspaces.

Check your AI systems, free

What the app says it does

What breaks the AI Act or GDPR in your code and logs right now, where and what to do, checked every hour and after each push; from code alone it shows possible problems (a chatbot with no AI notice, personal data sent to a model, a decision about people with no review), each marked Possible; every open problem is a Fix step on your plan; a system no check reads is shown as Not checked, never as OK; a problem that is gone at the next check is marked fixed, and a confirmed audit answer it contradicts is marked Needs review, never rewritten, put in Review and named in one letter to the admins from the same check.

Illustration · not your dataLaw Article 50(1) · AI ActProblem Possible · a chat answers people with no AI noticeWhere GitHub · web/chat.tsx, line 42 · See fix

The six checks

In logs (Datadog, Azure Monitor or a log system sending to a signed address): the first message to a person carries no AI notice; personal data such as an email address, a phone number or a card number in a prompt; a decision about a person with no human review, where the log shows it.
In code (GitHub, GitLab or the scanner in your own CI): the same three, each titled Possible and badged Possible, from code.

What it shows

A headline from the rows on the page, how many systems were checked and when, and one row per open problem: the provision, the problem, the file and line or the service, and what to do.
See fix opens the explanation with a code example. Mark as fixed and Not a problem, with a reason, are kept.
A system no check reads is listed as Not checked, never as OK; code alone never makes a system OK.
A problem that a later read could have seen and did not is marked Fixed.

What it does with a problem

Every open problem becomes a step in your plan, under Fix what Live found.
A confirmed audit answer that an open problem contradicts is marked Needs review and put in Review without being rewritten; the admins get one email.

What it does not do

It does not decide that the law is broken. A finding from code is a possibility, and none of the code rules has a measured precision yet, so none of them is mailed.
It reads only what your connections reach, and only since the last read, at most one day back for logs.

Which plan

Growth, which is not for sale yet: Live is open today to pilot workspaces with a connection. Starter shows what Live does, locked.

Where it sits in the law

Article 50(1) of Regulation (EU) 2024/1689: a provider designs an AI system intended to interact directly with natural persons so that they are informed they are interacting with an AI system, unless that is obvious to a reasonably well-informed, observant and circumspect person given the circumstances and context of use. It does not apply to systems authorised by law to detect, prevent, investigate or prosecute criminal offences, unless the public can use them to report one.
Articles 14 and 26: human oversight of high-risk AI systems, designed in by the provider (Article 14) and assigned by the deployer to people with the necessary competence, training, authority and support (Article 26(2)). Live flags a missing review on any decision about people; the duty applies where the system is high-risk.
Article 5(1)(c) of the GDPR: personal data adequate, relevant and limited to what is necessary in relation to the purposes for which they are processed. Live flags personal data in a prompt; whether it is necessary for your purpose is your call.

Starter

€499 a month

Prices exclude VAT.

Up to 20 AI systems
Evidence and the audit log
Your first system is free, with no end date

Check your AI systems, free

All plans

Questions

Does Live tell us we break the AI Act?

No. It shows where a duty may not be met, with the provision it comes from and what to do. A finding from code is always marked Possible, a person decides what is a problem, and Not a problem is kept with its reason. Nothing is sent to an authority.

Why is a system shown as Not checked?

Because no connection reads it. A system counts as checked only when a log connection is linked to it, or a check already found something on it. Code alone never shows a system as OK, so a quiet result is never mistaken for a clean one.

How often does Live run?

Every hour, and after each push to the default branch through the GitHub App or a GitLab webhook you add. Log reads look only at events since the last read, at most one day back. Excerpts are masked, at most three per problem, and deleted after 30 days.

Which plan includes it?

Growth, which is not for sale yet; today Live is open to pilot workspaces with a connection. Starter, at €499 a month for up to 20 AI systems, shows what Live does, locked. Prices exclude VAT.