AI Terms | HeadHonta

1. Overview

These AI Terms describe how HeadHonta builds, deploys, and governs artificial intelligence features across the platform. They supplement — and where applicable form part of — our Terms of Service, Privacy Policy, and Data Protection Document. By using HeadHonta's AI features you agree to these terms.

Our goal is simple: deliver AI that makes recruiters faster without replacing their judgment, and that respects the rights of the candidates whose data we process on our customers' behalf.

2. How HeadHonta Uses AI

HeadHonta uses AI as a productivity tool, not as a decision maker. We use large language models (LLMs) and embedding models from established providers, accessed through their APIs. We do not train foundation models ourselves; we orchestrate them with carefully designed prompts, retrieval, and guardrails.

Every AI-generated output is treated as a suggestion that a human reviews before a hiring action is taken.

3. AI-Powered Features

The current AI-powered features include:

CV parsing — extracting structured data (skills, roles, dates, education) from uploaded CVs.
Role-fit scoring — comparing a candidate against the job's defined requirements and producing a score with a written rationale.
CV review — an AI-assisted feedback report for candidates on the strength of their CV.
Candidate sourcing — ranking and summarising candidates from the customer's pool against an open role.
Interview assistance — live transcription, question suggestions based on the JD/CV, and note-taking via the browser extension.
Email drafting — templates and drafts for outreach, scheduling, and follow-up.
Summarisation — concise summaries of candidate profiles, interviews, and pipeline activity.

4. AI Providers & Sub-processors

HeadHonta routes AI inference through OpenRouter, an AI gateway that forwards prompts and embedding requests to underlying model providers (for example, the configured large language and embedding models). Interview transcription is handled by a self-hosted speech-to-text service (faster-whisper) within our infrastructure, not a third-party provider. OpenRouter is listed as a sub-processor in our Data Protection Document.

5. Training, Retention & Confidentiality

HeadHonta does not use customer data or candidate data to train its own models.
We request that our AI providers do not use customer prompts or outputs to train their foundation models. Because OpenRouter routes requests to underlying model providers, whether this is contractually guaranteed depends on the specific provider serving a given request.
Provider-side retention of API content is governed by OpenRouter's terms and those of the routed provider; details can be shared on request.
Interview audio transcribed by our self-hosted speech-to-text service is processed in memory and persisted only as the transcript text returned to the customer's workspace.

6. What Data is Sent to AI Models

The data we send to AI models is the minimum needed to perform the requested feature, for example:

CV text or extracted fields for parsing and scoring.
Job description text for matching and scoring.
Conversation context for email drafting and summarisation.
Live interview audio (streamed to the self-hosted STT service) and the resulting transcript for interview assistance.

We do not send protected characteristics, demographic data, government identifiers, or payment data to AI models as scoring inputs.

7. Human-in-the-Loop Decisions

HeadHonta does not make hiring decisions. AI outputs are recommendations that a recruiter or hiring manager reviews before any action is taken. Customers configure their own workflows, stages, and approvals; HeadHonta does not automatically reject, hire, or progress candidates without an authorised human action.

8. Fairness, Bias & Non-discrimination

We design AI features to evaluate candidates on the skills, experience, and qualifications defined for the role, not on demographic attributes. Specifically:

Role-fit scoring uses the requirements that the customer defines for the job.
The platform does not infer or use protected characteristics (race, ethnicity, gender, religion, age beyond legally relevant minimums, disability, sexual orientation) as scoring signals.
Every AI score is accompanied by a written rationale so recruiters can review the reasoning and override the recommendation.
We periodically review prompts and evaluation logic for language that could introduce bias.

Customers remain responsible for ensuring that their use of HeadHonta complies with applicable employment, equality, and non-discrimination laws in their jurisdiction.

9. Limitations & Hallucination Risk

Like all LLM-based systems, HeadHonta's AI features can occasionally produce inaccurate or fabricated content ("hallucinations"). Customers should:

Treat AI outputs as draft material to be reviewed, not as facts.
Verify critical claims (employment history, qualifications, compensation, identity) through authoritative sources.
Use HeadHonta's human-review interfaces before sending AI-drafted communications to candidates.

10. Customer Responsibilities

By using AI features, the customer agrees to:

Upload only data they have a lawful basis to process and share with a service provider.
Provide candidates with appropriate notice and, where required, obtain consent for AI-assisted processing.
Make the final hiring decision and not rely solely on AI outputs.
Use HeadHonta in compliance with applicable laws, including the EU AI Act, GDPR, UK GDPR, NDPA, CCPA/CPRA, and local employment law.
Not attempt to reverse-engineer, extract weights from, or misuse the AI features (including prompt injection intended to exfiltrate other customers' data).

11. Candidate Rights

Candidates whose data is processed through HeadHonta are entitled to:

Be informed when AI is used to assist in evaluating their application.
Request a human review of an AI-assisted assessment from the customer organisation.
Exercise the data subject rights described in our Privacy Policy and GDPR statement.

12. Opt-Out & Disabling AI Features

Customers can disable AI-powered scoring and AI-drafted communications on their account by contacting sales@headhonta.com or their account manager. Some features (e.g., CV parsing) are foundational to the service and cannot be fully disabled without losing core functionality; alternative manual workflows are available for sensitive cases.

13. Security of AI Workloads

All traffic to AI providers is encrypted in transit.
API credentials for AI providers are stored in managed secret stores and rotated periodically.
Inference requests are scoped to the minimum necessary data and tied to the originating tenant for auditability.
Outputs are stored within the customer's tenant and subject to the same access controls as the rest of the workspace.

14. Changes to These AI Terms

We may update these AI Terms as our features, providers, or applicable law change. Material changes will be communicated by email to active customers or by an in-product notice at least 30 days before they take effect.

15. Contact

Questions about our AI features or these AI Terms can be sent to:

AI Terms & Conditions