The AI Act introduces regulations for artificial intelligence systems by classifying them according to their level of risk. Depending on their use, AI systems may be prohibited, subject to restrictions or, as in the case of recruitment, classified as high-risk systems.
As Hero Parser develops and as the interpretation of the regulations evolves, we assume that the system qualifies as a high-risk system and is developed in accordance with the requirements applicable to this category.
Is Hero Parser a high-risk system?
Yes.
Systems used to assess, select and support recruitment decisions have been clearly classified under the AI Act as high-risk systems.
Hero Parser analyzes candidates’ CVs and supports recruiters in assessing their fit for a specific role, which means it is subject to the requirements provided for this category.
What does this mean in practice?
Being classified as a high-risk system does not mean that the tool is dangerous. It means that it must meet a number of requirements related to:
- transparency of operation
- risk control
- data quality
- human oversight
- auditability
Hero Parser has been adapted to these requirements both at the technical and organizational level.
How does Hero Parser meet the requirements of the AI Act?
a) Risk management
We have implemented a risk management system that includes the identification, analysis and mitigation of potential risks related to the use of AI in recruitment.
We analyze, among other things:
- the risk of incorrect CV interpretation
- the risk of unintended bias in assessment
- the risk of incorrect candidate matching
The system is continuously monitored, and the results of these analyses are used to improve it.
b) Data management
Hero Parser processes data provided as part of the recruitment process.
Before AI is used to analyze a CV, the document may be anonymized by removing identifying data such as first name, last name or contact details.
We do not use external data without the user’s knowledge. Candidate data is stored only for the duration of the recruitment process.
c) Transparency
ZWe ensure transparency in how the system works:
- each candidate assessment can be broken down into the factors that influenced the score
- the user can see what the assessment is based on
- the system clearly communicates that it uses AI
As a result, decisions made with the support of Hero Parser are understandable and can be justified.
The recruiter can always see what the fit score of a specific CV against a job offer is based on and can correct it.
d) Documentation
We maintain technical documentation of the system, covering:
- how the algorithms work
- the scope of data used
- quality metrics
- test results
The documentation is updated on an ongoing basis and is available for audit purposes.
e) Monitoring and accountability
The system is continuously monitored in terms of performance quality and consistency of assessments.
We have introduced versioning of models and algorithms, which means that every decision can be reproduced in the context of a specific version of the system.
f) Human oversight
Hero Parser does not make recruitment decisions independently.
The system presents recommendations, while the recruiter can:
- accept them
- modify them
- reject them
The final decision always belongs to a human.Ostateczna decyzja zawsze należy do człowieka.
g) Decision logging and auditability
We provide the ability to reproduce the candidate assessment process by recording:
- input data
- analysis results
- decision context
This makes it possible to conduct an audit and explain how the system operated in specific cases.
h) Secure infrastructure
The system is secured in accordance with current security standards, including:
- encrypted communication
- protection against unauthorized access
- regular backups
The infrastructure meets high data protection requirements.
Summary
Hero Parser is a high-risk system within the meaning of the AI Act and has been designed to meet the requirements provided for this category.
Instead of avoiding this classification, we have adopted it as a starting point for building a system that is:
- transparent
- controllable
- secure
- auditable
We continuously monitor regulatory developments and adapt the system to applicable legal requirements.
