The situation
A description of an AI feature and how customers will use it
A recruiting-software company is adding a feature that ranks applicants against a job description, built on a general-purpose model from a third-party provider through an API. Its customers are employers across Europe. The product team plans a beta in two months and has asked counsel what the AI Act means for them.
Counsel has to answer in order: is the feature an AI system, who is the provider and who the deployer, does ranking candidates make it high-risk, and what follows for the company and its customers. The duties are long, and their application dates have been revisited, so the note must say what applies and when.
What the research has to settle
The rules behind the answer
- Provider or deployer
- Under the AI Act, the provider develops an AI system, or has it developed, and places it on the market under its own name. The deployer uses it under its own authority. The recruiting company is the provider of the feature, and its customers are deployers.
- Recruitment is a listed high-risk use
- AI systems intended for recruitment or selection, including filtering applications and evaluating candidates, are among the high-risk uses listed in the Act, subject to its narrow exceptions.
- High-risk providers carry the main duties
- Providers of high-risk systems need risk management, data governance, technical documentation, logging, human oversight, accuracy and security measures, a conformity assessment and registration. Deployers have their own duties, including human oversight and informing workers.
- The model provider is a separate role
- The company supplying the general-purpose model has its own obligations under the Act. Building on its model does not move the feature’s duties to it. When each set of duties applies must be confirmed on the date of the note.
How it runs in azta
From the question to the document
The question, as you would type it
We sell recruiting software in the EU and are adding a feature that ranks applicants against a job description, using a third-party general-purpose model through an API. Under the AI Act, are we the provider, is the feature high-risk, what must we do before release, and what must our customers do?
You start with
A description of an AI feature and how customers will use it
- 01
Describe the feature
Give what the feature does, which model it uses, who uses the output and for what decision. azta searches the AI Act and related EU law.
Cited research· Free - 02
Read the definitions and annex
Open the definitions of provider and deployer and the high-risk annex in the reader, with their in-force status on the date you ask about.
Atlas statute reader - 03
Draft the internal note
Ask for a note on role, risk category and duties, with a release checklist. Edit it in Canvas.
Canvas · Ask azta· Pro - 04
Keep it with the product team
On Enterprise, keep the note and research in a shared workspace with product and engineering, so later questions start from the same file.
Workspaces· Enterprise You send
An internal note on the company’s AI Act role, risk category and duties
What you get
An internal note on the company’s AI Act role, risk category and duties
Internal note · draft
Candidate ranking feature: AI Act position
Our role. We place the ranking feature on the market under our name, so we are its provider. Customers who use it in hiring are deployers. The model supplier is a provider of a general-purpose AI model and has separate obligations.
Risk category. Ranking applicants for a vacancy is a recruitment use listed as high-risk. None of the narrow exceptions fits a feature whose output shapes which candidates a recruiter sees, so we should plan on the high-risk regime.
Before release. We need the risk management, data governance and documentation work, a human-oversight design, the conformity assessment and registration, and instructions for deployers. The annex sets out when each duty applies, as confirmed on the date of this note.
Outcomes on your numbers
What azta is worth on this task
Pro pays for itself after
13 minutes a month
At €130 an hour, that is all the time Pro (€27 a month) has to take off your desk to cover its price.
Example figures. Replace them with your own.
- Hours back a year
- 25
- Capacity at your rate
- €3,276
- More AI Act notes a quarter
- 0.7
- Return on Pro
- 10.1×
- Research
- Drafting
- Your review
A scenario you set, not a promise. The hours that come back go to another client, or deeper into this one.
Want these numbers on your team’s real workload?
Book a demoWhat stays with you
- The technical design and documentation of the feature.
- The judgement on whether any exception applies.
- The release decision.
Frequently asked questions
- Does azta keep up with changes to the AI Act?
- Legislation is indexed daily, and each citation shows its in-force status. Amendments still under negotiation are not law yet and need following separately.
- Does the GDPR still apply?
- Yes. Ranking candidates involves personal data, so the GDPR applies alongside the AI Act, including its rules on automated decisions. Ask about both.
- Is our product description used for training?
- No. Client documents and questions are never used for training on any plan, and they stay in the EU/EEA.
