CometHire AI & Data Trust Center

Purpose of AI in CometHire

CometHire uses AI to support structured analysis of application documents and evidence-based candidate review. Our AI features assist recruiters during screening and evaluation while keeping people fully in control of every hiring decision.

AI-supported workflows

CometHire uses large language models to analyse job descriptions and application documents in order to support structured candidate evaluation. AI-supported features may help users:

• generate and structure evaluation criteria
• highlight relevant information within application documents
• provide requirement-level evidence assessments (Limited, Partial, Good or Strong)
• aggregate those assessments into one overall evidence level per candidate
• suggest interview or reference check questions
• summarise candidate qualifications and experience
• generate structured candidate reports

All AI-supported outputs are intended to support human review and interpretation. Recruiters remain responsible for defining evaluation criteria, reviewing candidate information and making final decisions.

Human oversight

CometHire is deliberately designed as a human-in-the-loop system. The software does not automatically reject or select candidates, and it does not tell recruiters whom to hire.

For each requirement a recruiter defines, CometHire shows how much supporting evidence it found in the application documents. Evidence is presented as one of four qualitative levels — Limited, Partial, Good or Strong — and is linked to the passage in the application it came from.

CometHire then aggregates these requirement-level assessments into a single overall evidence level per candidate, using a fixed rule and the requirement weightings the recruiter has defined. Candidate lists can be sorted by that level.

What this level says is how well an application evidences the requirements the recruiter defined. It is not a judgement about the person and not a hiring recommendation. Two candidates at the same level can be very different, and the evidence behind every assessment is one click away.

Scoring is done by people. Recruiters rate requirements themselves and set the weightings; until a recruiter enters a rating, the score field is empty. CometHire never fills it in.

Advancing or rejecting a candidate is an individual action taken by a recruiter, one candidate at a time. There is no bulk advance or bulk reject, and nothing happens automatically because of an evidence level.

Users can review, adjust, ignore or overwrite AI-supported outputs at any time.

How an application is processed

1. The application document is received and converted into a text transcript.
2. The transcript is de-identified to reduce personally identifying information.
3. The de-identified transcript, together with the requirement profile defined by the recruiter, is analysed to produce requirement-level evidence assessments.
4. Those assessments are aggregated into one overall evidence level per candidate, using a fixed rule and the recruiter's weightings.
5. Results are presented to the recruiting team, which reviews the evidence and decides.

The model that carries out the evaluation in step 3 never receives the candidate's name or contact details. The de-identification in step 2 is itself performed by an AI provider, which does receive the application in its original form in order to remove that information. Both providers are named below and are established in the European Union.

Infrastructure and AI providers

CometHire processes data exclusively within EU-based infrastructure.

The AI models used for evaluation are operated by Amazon Web Services within the region stated above. The model developer does not receive the data. Prompts and outputs are not made available to the model developer and are not used for model training.

No personal data processed for our customers is transferred outside the European Union.

AI training data

CometHire does not use customer data or candidate data to train AI models. Your candidate and hiring data remain private and are never fed back into any model or used for general model training. Our AI providers are contractually excluded from using this data to train or improve their own models.

Transparency and user control

AI-supported features are marked in the product with a sparkle icon, so users can see when a feature is driven by generative AI. Most AI functionality is optional and can be used flexibly depending on the organisation's hiring process.

Where the law stands

The EU AI Act entered into force in August 2024 and applies in stages. As of September 2026:

• In force since February 2025: the prohibitions on unacceptable-risk AI practices (Article 5) and the AI literacy obligation (Article 4).
• Applicable from 2 August 2026: the transparency obligations under Article 50, including informing people when they interact with an AI system.
• Deferred: the obligations for stand-alone high-risk AI systems under Annex III — which covers AI used to analyse and filter job applications and to evaluate candidates. Under the Commission's Digital Omnibus package, the application date for these obligations has been moved from 2 August 2026 to 2 December 2027. This change takes legal effect on formal adoption and publication in the Official Journal.

CometHire falls within the Annex III category described above. We follow these developments and assess how the obligations apply to our product and to our responsibilities as a provider of AI software. We do not claim to have completed a conformity assessment for obligations that are not yet applicable.

How we work with the Act's principles

Human oversight.
CometHire supports human decision-making rather than replacing it. Recruiters define the evaluation criteria and the weightings, review the outputs and make every final decision.

Transparency and explainability.
Evidence assessments are linked to the passages in the application they are based on, so a user can check the basis for an assessment rather than trusting a summary. The overall evidence level can be traced back to the requirement-level assessments it was calculated from.

Privacy and data governance.
Data minimisation, de-identification before evaluation, defined retention periods, and EU-based infrastructure and AI providers.

Fairness and non-discrimination.
Our platform terms do not permit discriminatory evaluation criteria. De-identification before evaluation is designed to reduce the influence of information that is irrelevant to the role.

Technical robustness and security.
Logging of processing steps including prompt and model versions, access controls, and monitoring of system performance.

Prohibited practices

CometHire does not develop or operate AI systems in the unacceptable-risk category under the EU AI Act. We do not use AI for social scoring, biometric categorisation based on sensitive attributes, emotion recognition, manipulative targeting, or fully automated hiring decisions.

What we ask of our customers

The EU AI Act places obligations on the organisations that deploy AI systems, not only on providers. Under our Terms of Service, customers agree to have a competent person review outputs before any decision affecting a candidate, to define requirement profiles that relate to the actual requirements of the role, to inform candidates where required, and to ensure the people using the product understand it well enough to use it appropriately.

Customers also agree not to use the overall evidence level as an automatic cut-off, and to review the underlying evidence before deciding. We provide candidate information notices and documentation to support this.

CometHire processes personal data on behalf of its customers in a secure, transparent and documented manner. Our approach includes data minimisation, defined retention periods, explainable AI-supported workflows, human oversight, and exclusively EU-based infrastructure and AI providers. Candidate data is not used to train AI models.

Roles and responsibilities

CometHire acts as a data processor across its product features. Our customers remain the data controllers and determine the purposes and means of processing candidate data.

Processing takes place solely on the documented instructions of the controller, under a Data Processing Agreement pursuant to Art. 28 GDPR, available at comethire.com/dpa. This applies consistently across AI-supported and non-AI features — using AI functionality does not change the underlying roles.

Data retention

Candidate data, including application documents, AI-supported analyses, generated reports and associated audit records, is retained only as long as necessary for the recruitment process. By default it is deleted six months after a recruitment process has been closed. This period supports legitimate recruitment purposes, including documentation of hiring decisions and the handling of potential legal claims. Customers may implement different retention periods where required by law or internal policy.

Technical processing and diagnostic logs are deleted after seven days.

Article 22 GDPR and automated decision-making

CometHire's AI-supported features do not constitute solely automated decision-making within the meaning of Article 22 GDPR.

CometHire supports human decision-making. Evidence assessments, summaries and insights are presented to recruiters or hiring teams for review and interpretation. CometHire does not produce suitability scores or hiring recommendations.

CometHire does produce an overall evidence level per candidate and allows candidate lists to be sorted by it. This constitutes profiling within the meaning of Art. 4 (4) GDPR. It does not constitute a decision within the meaning of Article 22, because no decision follows from it: advancing or rejecting a candidate is an individual action taken by a person for each candidate, and no action in the product is triggered by an evidence level or by any calculated value.

Customers should use the overall evidence level to organise review, not as a cut-off that decides the outcome by itself. Final decisions remain fully under the control of the customer.

Candidate information and customer responsibilities

CometHire's workflows process candidate application documents, including CVs. Candidates should therefore be informed before their documents are uploaded or processed.

Where candidates apply through a CometHire-powered application form, information about AI-supported processing is presented before submission. Once an application is submitted, it may immediately enter the parsing and evaluation workflow described in Section 1.

CometHire relies on its customers — employers, recruiting teams and integration partners — to provide the appropriate notices and to establish any required lawful basis before candidate data is uploaded into the system. Section 5 of this page is written for candidates and can be linked directly.

DPIA and compliance support

We support customers in conducting Data Protection Impact Assessments and other internal compliance assessments. Customers remain responsible for determining the lawful basis for processing candidate data; we provide transparency into how our systems function, which data is processed, which AI-supported features are used and which safeguards are in place. Additional technical and compliance documentation is available on request.

De-identification before evaluation

Before an application is evaluated, CometHire applies an AI-assisted de-identification step intended to reduce personally identifying information, including names, contact details, age-related information, gender indicators and profile images where detected.

The goal is to support evaluations focused on skills and qualifications. Contextual information relevant to assessing professional qualifications — previous employers, educational background, project experience — may remain part of the analysis where it is needed for the recruiting workflow.

Because the process is AI-assisted, complete removal of all identifying or sensitive information cannot be guaranteed in every case.

Blind view for recruiters

Separately from the de-identification step above, recruiters can choose to hide candidate details in their own view of the application during early-stage review. This is a display setting controlled by the user; the de-identification before evaluation happens regardless.

Structured evaluation

CometHire assesses candidates against role-specific criteria defined in advance rather than against unstructured impressions. Recruiters define and weight the criteria based on the requirements of the role.

The system analyses application documents against those criteria, links its evidence assessments to the relevant passages, and aggregates them into one overall evidence level per candidate using a fixed rule and the recruiter's own weightings. That level describes how well an application evidences the defined criteria. It is not a judgement about the person, and the assessments behind it remain visible.

Combining defined criteria with evidence-linked analysis is intended to reduce ambiguity and support more consistent evaluation.

Ongoing evaluation and safeguards

We continuously evaluate and refine our AI-supported workflows to improve transparency, explainability and consistency across hiring scenarios. Our approach combines selected third-party AI models with our own evaluation logic, prompting strategies, de-identification workflow and product safeguards.

This section is written for candidates and can be linked directly.

The company you are applying to uses CometHire to support parts of its recruitment process. CometHire is software that helps recruiters review applications in a more structured and consistent way. Some features use artificial intelligence to analyse application documents and support recruiters during review.

How AI is used

AI may be used to analyse the information in your application documents, identify skills, qualifications and experience relevant to the role, assess how well your application evidences each requirement, generate summaries for recruiters, and suggest interview questions based on the role's requirements and your application.

The purpose is to help recruiters review applications more consistently.

What AI does not do

The AI does not decide anything about your application. It does not hire, it does not reject, and it does not tell recruiters who should get the job. It does not give you a score.

What it does is check your application against the requirements the employer defined, show how well each one is evidenced, and summarise that into an overall level. Recruiters can sort applications by that level to organise their review — and they can open the evidence behind every single assessment.

Any score you may see or hear about was given by a recruiter, not by the AI.

Every decision to move an application forward or to close it is made by a person, one application at a time.

Human review and oversight

Recruiters remain responsible for evaluating candidates and making all final decisions. AI-supported outputs support human judgement rather than replacing it, and recruiters can review, interpret, ignore or challenge them at any time.

How we work to reduce bias

De-identification before analysis.
Before AI-supported evaluation takes place, your application document goes through a de-identification process intended to reduce personally identifying information such as your name, contact details, age and gender. The model that evaluates your application against the role's requirements does not receive that information. The goal is to focus the evaluation on qualifications, skills and job-related experience.

Structured evaluation.
You are assessed against requirements defined in advance for the role, rather than against unstructured impressions.

Evidence-based outputs.
For each requirement, the system indicates the level of supporting evidence it found in your application — Limited, Partial, Good or Strong — and links it to the passage it came from. These levels are also summarised into one overall level for your application. Recruiters can review the underlying evidence rather than relying on a summary or a classification.

How your data is processed

Your application data is processed only for recruitment purposes and is not used to train AI models. It is processed exclusively within the European Union.

Your application data, including documents, AI-supported analyses, generated reports and associated audit records, is retained only as long as necessary for the recruitment process. By default it is deleted six months after the recruitment process has been closed, unless the hiring organisation or applicable law requires a different period.

Your rights

Depending on applicable law, you may have rights relating to your personal data, including access, rectification, deletion, restriction of processing and objection.

The organisation you applied to decides how your application is reviewed and makes all hiring decisions. It is the controller for your data; CometHire processes information on its behalf. To exercise your rights or to ask how your application was handled, please contact that organisation.

For questions about CometHire and our approach to responsible AI, see the sections above or write to anja@comethire.com.