Responsible AI & Governance
Resourcehip uses AI to produce HIP Scores — product ratings based on material resilience, repairability, and environmental impact. This page sets out the voluntary safeguards we apply and how we stay accountable.
AI Act status
The EU AI Act (Regulation 2024/1689) entered into force on 1 August 2024, with obligations phased in over several years. High-risk AI obligations under Annex III apply from 2 August 2026.
The HIP Score is not a high-risk AI system under the EU AI Act. High-risk classification under Article 6(2) applies to AI systems listed in Annex III: biometrics, critical infrastructure, education, employment, credit scoring, law enforcement, migration, and administration of justice. A consumer product sustainability rating is not in that list.
As a UK-based provider, the AI Act may apply to Resourcehip where our output is used in the EU (Article 2). Our most likely classification is limited or minimal risk, where the main obligations are AI literacy (Article 4) and, where relevant, transparency (Article 50).
We apply the safeguards described below as voluntary good practice, not because a regulator requires them for our risk classification.
What we do
These governance practices go beyond what is legally required for a system at our risk level.
- Documented risk assessment. We maintain a risk register covering category bias, data quality, model drift, and dispute backlogs. Read the full Risk Register.
- Performance testing. Before launch and periodically, we test for bias across product categories, robustness with incomplete data, and out-of-distribution behaviour. Results are published on our Audit Reports page.
- Human review before publication. No rating is published automatically. Every HIP Score is reviewed and approved by a human before it appears on the website.
- Published dispute and appeal processes. If you believe a score is based on incorrect data, you can request a formal review. See Dispute a Rating. For a formal appeal, see the Appeal Procedure.
- Ongoing monitoring. We re-test reference products monthly and conduct quarterly bias audits to detect drift or systematic issues.
How our AI pipeline works
HIP Scores are produced using a local AI pipeline running on Resourcehip's own hardware. No product data is sent to external cloud AI services.
| Component | Detail |
|---|---|
| AI model | qwen3.5:35b, running locally via Ollama |
| Scoring method | Structured rubrics across 7 dimensions — not a pure neural black-box |
| Data sources | Public, peer-reviewed datasets (USGS, EPA, LCA databases) |
| Temperature | 0.1 — produces more consistent outputs |
| Human oversight | Every rating reviewed before publication |
For full scoring methodology details, see the Methodology page.
Risk management
We have identified and documented the following risks:
| Risk | Mitigation |
|---|---|
| Category bias — model under/over-scores a product category | Conservative category defaults; quarterly bias audit; rubric-based scoring |
| Data quality / hallucination — model produces scores contradicting input data | Rubric-based scoring; mandatory human review; conservative defaults |
| Model drift — scores change unexpectedly after model updates | Version-pinned AI model; re-validation required before any update |
| Dispute backlog — challenges not resolved in time | Published dispute and appeal SOPs; 30-day response SLA; CEO escalation path |
Read the full Risk Register.
Transparency and accountability
- AI disclosure. HIP Scores are produced by an AI scoring pipeline. This is disclosed on every rating page and on our Methodology page.
- Published methodology. Our scoring methodology, dimension weights, and data sources are fully published. See Methodology.
- Score challenges. If you believe a rating contains a factual error, see Dispute a Rating. For a formal appeal, see the Appeal Procedure.
- Risk register. The full Risk Register is published on this site.
Complaints and regulatory contacts
Score disputes: If you believe a rating contains a factual error, see Dispute a Rating or email methodology@resourcehip.com.
Formal appeals: For a formal appeal with independent review escalation, see the Appeal Procedure or email score-appeal@resourcehip.com.
UK data protection: If you have a concern about how Resourcehip handles personal data, you may contact the UK Information Commissioner's Office (ICO) at ico.org.uk.
EU AI Act: If you believe the AI Act applies to your use of HIP Scores and wish to raise a concern, the relevant authority is the market surveillance authority in your EU member state (Article 85 of Regulation 2024/1689). The ICO does not enforce the EU AI Act.
Audit reports
We publish periodic audit reports covering bias testing, score distribution analysis, and dispute outcomes. See Audit Reports.
Regulatory references
- EU Regulation 2024/1689 (EU AI Act) — Articles 2, 4, 6, 50, 85
- Risk Register
- Resourcehip Methodology
- Dispute a Rating
- Appeal Procedure
- Audit Reports
Provider: Resourcehip Ltd (SC873386), Scotland
Page last updated: 16 June 2026
Next review: 16 December 2026
Questions about our AI governance? Email hello@resourcehip.com.