The Pathwize Blog

Notes on verifiable AI data and EU compliance

Notes on verifiable AI data and EU compliance

PathwizeThe verifiable data layer for AI

Guides

The best providers for expert SFT and domain data in 2026

Where to source gold answers and expert judgments for fine-tuning, ranked by author qualifications, nuance, and provable provenance.

July 16, 2026

Guides
GuidesPathwize AI
The best providers for AI red teaming in 2026
Who to trust to break your model before your users do, ranked by domain depth, reproducibility of findings, and auditability.
July 14, 2026
Guides
Pathwize AIGuides
The best providers for RLHF and preference data in 2026
Where to source human preference data, ranked by what protects the reward signal: real expertise, defence against AI-in-the-loop, and provenance.
July 10, 2026
Guides
GuidesPathwize AI
The best providers for model evaluation in 2026
A buyer's guide to the options for human model evaluation, ranked by the criteria that actually matter: expert quality, agreement, provenance and EU fit.
July 8, 2026
Guides
GuidesPathwize AI
How to evaluate an AI data vendor: a due-diligence checklist
The questions that separate a data vendor you can defend to an auditor from one that will cost you a re-labelling project six months in.
July 4, 2026
Provenance
ProvenancePathwize AI
A provenance trail every AI auditor will accept
Auditors do not want a summary, they want to follow one output back to the person and process that produced it. Here is what a trail they trust looks like.
July 2, 2026
Guides
GuidesPathwize AI
Domain judgment and SFT tasks: expert answers as training data
Sometimes you do not want a score or a comparison, you want the right answer, written by someone qualified to write it. How domain judgment tasks produce gold data.
July 2, 2026
Experts
InsightPathwize AI
Experts
How to become an AI data expert: a step-by-step guide
If you have real domain expertise, AI labs will pay for your judgment. Here is how the work actually works and how to start without getting scammed.
July 1, 2026
Compliance
Pathwize AICompliance
The EU AI Act Annex IV checklist for your training data
A field-tested checklist for the training-data documentation the EU AI Act expects from high-risk systems, and how to produce it without slowing your team down.
June 30, 2026
Experts
ExpertsPathwize AI
How to get paid as a domain expert for AI training work
Your professional expertise is now valuable to AI teams. Here is what expert AI data work is, what it pays, and how to start without leaving your field.
June 28, 2026
Guides
Pathwize AIGuides
How to run a human-data pilot before you commit
A pilot is the cheapest way to learn whether a data partner can actually do your hardest work. Here is how to design one that tells you something real.
June 27, 2026
Experts
ExpertsPathwize AI
How lawyers earn with AI data work
Legal AI needs real legal judgment. Here is how qualified lawyers contribute flexibly, and get paid for the expertise they already have.
June 26, 2026
Guides
GuidesPathwize AI
Red teaming tasks: stress-testing a model for the failures that matter
Normal grading measures typical behaviour. Red teaming hunts for the inputs that break the model. How adversarial tasks work and what they produce.
June 25, 2026
Research
InsightPathwize AI
Research
Live inter-rater agreement: catching data drift as it happens
Waiting for eval regressions to reveal bad data is too late. Live agreement between reviewers flags drift while you can still fix it.
June 24, 2026
Experts
ExpertsPathwize AI
Scientists and researchers: how to earn from AI data work
Frontier models get stuck on exactly the problems scientists are trained for. Your judgment on hard technical questions is worth real money to AI labs.
June 24, 2026
Provenance
ProvenancePathwize AI
How to prove your RLHF data was actually labelled by humans
AI quietly routed through a contractor looks like human feedback until it poisons your reward model. Here is how to detect it and design a pipeline that can prove the human signal.
June 22, 2026
Research
InsightPathwize AI
Research
What is model evaluation? A practical guide for AI teams
Benchmarks tell you how a model does on a fixed test. Evaluation tells you whether it is good enough for your users. The difference matters more than it sounds.
June 20, 2026
Experts
ExpertsPathwize AI
What is RLHF work, and how do specialists do it?
RLHF is how AI models learn from human feedback. Here is a plain-English explanation and how experts can get involved and paid.
June 18, 2026
Guides
GuidesPathwize AI
Evaluation and grading tasks: scoring model outputs against a rubric
When you need a number you can track, you need an evaluation task. How rubrics turn a vague sense of quality into a measurement graders agree on.
June 18, 2026
Experts
ExpertsPathwize AI
Finance and accounting experts: earn with AI data work
AI is confidently wrong about numbers, regulation and risk in ways only a finance professional spots. That gap is what labs pay you to close.
June 17, 2026
Compliance
CompliancePathwize AI
Article 10 data governance in plain English
Article 10 of the EU AI Act asks for data governance on your training, validation and testing sets. Here is what that means without the legalese.
June 16, 2026
Compliance
CompliancePathwize AI
GDPR-safe AI training data: an EU-native approach
Sourcing human data for AI without a GDPR headache means keeping data in the EU, minimising what you collect, and proving lawful processing end to end.
June 14, 2026
Research
InsightPathwize AI
Research
Gold-standard datasets: how to build one your team trusts
A gold set is the reference your whole quality process leans on. If it is quietly wrong, everything measured against it is quietly wrong too.
June 13, 2026
Experts
ExpertsPathwize AI
How software engineers earn reviewing AI code
Coding models need engineers who can spot the subtle bug an eval will miss. Here is how developers get paid for exactly that.
June 12, 2026
Guides
InsightPathwize AI
Guides
Preference and RLHF tasks: teaching a model what good looks like
Pairwise comparison is the workhorse of RLHF. How preference tasks work, why the rationale matters as much as the choice, and how to design them well.
June 11, 2026
Compliance
CompliancePathwize AI
EU data residency for AI teams, without the friction
Keeping data in the EU is the single biggest de-risking move for European AI, and it does not have to slow you down.
June 10, 2026
Experts
Pathwize AIExperts
Do you pay tax on AI data work? A simple guide for EU contributors
Money from AI data work is income, and income is usually taxable. A plain guide to what that means for freelancers in the EU, without the jargon.
June 10, 2026
Experts
ExpertsPathwize AI
How clinicians earn with AI data work (without leaving medicine)
Medical AI needs real clinical judgment. Here is how doctors, nurses and allied health professionals contribute flexibly and get paid for it.
June 8, 2026
Compliance
Pathwize AICompliance
Data minimization and PII handling in AI training pipelines
The safest personal data is the data you never collected. A practical look at minimization, redaction and lawful handling for AI teams.
June 6, 2026
Research
ResearchPathwize AI
Expert evaluation vs crowd labelling for frontier models
Crowd labelling scales cheaply but breaks on hard, high-stakes tasks. Here is when expert evaluation pays for itself and how to combine the two.
June 5, 2026
Experts
Pathwize AIExperts
How much does AI data work pay in 2026?
Pay for AI data work varies widely by expertise and task. Here is how to think about rates, and why specialists earn more.
June 4, 2026
Guides
GuidesPathwize AI
The four types of AI data tasks, explained
Preference, evaluation, red teaming and domain judgment. What each task type does, when to use it, and how they fit together in a data programme.
June 4, 2026
Company
InsightPathwize AI
Company
The hidden cost of an in-house labelling team
An in-house data team looks cheaper until you count recruiting, classification, payments, tooling and the senior time it quietly consumes.
June 2, 2026
Experts
InsightPathwize AI
Experts
Is AI data work legit? How verified expert networks pay and protect you
AI data work is real and can pay well, but the marketplace is uneven. Here is how to tell a trustworthy platform from a risky one.
May 30, 2026
Research
ResearchPathwize AI
How to measure and reduce annotation bias
Bias in your labels becomes bias in your model. The good news is that most annotation bias is measurable, and much of it is fixable at the guideline.
May 30, 2026
Experts
InsightPathwize AI
Experts
What is AI red-teaming, and can experts do it?
Red-teaming is the craft of finding where a model fails before users do. Here is what it involves and how experts get paid to do it.
May 28, 2026
Provenance
InsightPathwize AI
Provenance
Data provenance for AI, explained
Provenance is the record of where your data came from and how it was produced. Here is why it is becoming the difference between a defensible model and a liability.
May 27, 2026
Research
ResearchPathwize AI
Human evaluation for reasoning models: what breaks
As models reason better, evaluation gets harder, not easier. Here is why generic raters fail on reasoning and what to do instead.
May 24, 2026
Experts
ExpertsPathwize AI
Your data rights as an AI data contributor in the EU
Contributing to AI should not cost you your rights. Here is what EU rules mean for how your data and work are handled.
May 20, 2026
Company
Pathwize AICompany
Buy vs build: sourcing human data for AI in the EU
Recruiting, classifying and managing experts in-house is slower and riskier than it looks. Here is an honest buy-vs-build comparison for European AI teams.
May 18, 2026
Compliance
CompliancePathwize AI
Worker classification for AI data work, done right
Misclassifying the people who produce your data is a legal and reputational risk. Here is how to get classification right across the EU.
May 16, 2026
Compliance
CompliancePathwize AI
What EU AI Act Annex IV actually requires of your training data
A practical read of the Article 10 / Annex IV provenance and traceability obligations, and what they mean for high-risk model builders.
May 12, 2026
Provenance
ProvenancePathwize AI
AI disguised as human feedback is poisoning RLHF
When contractors quietly route tasks through an LLM, the human signal you paid for disappears. How to detect and design against it.
April 28, 2026
Research
Pathwize AIResearch
Trust scores belong on data batches, not just people
Why per-batch trust and live inter-rater agreement beat after-the-fact spot checks for catching quality drift.
April 9, 2026

Latest updates

From the team