Fields where errors cost mostEach judged by people who practise it.

Models now answer questions people used to bring to professionals. In those fields, a confident wrong answer is a safety problem, not a style problem.

The people who can tell a correct answer from a plausible one are the people who do the work. That is why Saolabs organises its network by industry and vets each expert on tasks from their own field.

Each industry has its own page covering why it matters for AI safety, what experts do, the failures they catch and how they are vetted.

The professions listed are where we recruit, not a closed list. If a project needs a narrower specialism within a field, vetting can add assessments for it.

• Medicine and law: doctors, nurses, pharmacists, lawyers, paralegals and compliance specialists
• Finance: analysts, accountants and risk specialists
• Engineering: mechanical, electrical and civil engineers
• Software: software engineers, architects and security specialists
• Science and mathematics: physicists, chemists, biologists, mathematicians and statisticians
• Languages: native speakers, translators and linguists

The tasks change by industry; the method does not. Every expert is sourced, interviewed by AI voice agents, assessed on real-world work tests in their field, and matched to the right project.

What changes is the content of the test. A lawyer is assessed on legal reasoning, a statistician on analysis, a translator on translation. The standard is the same: can this person do the work the model needs to learn?

Across all fields, experts produce labelled data, expert-written training data, evaluations, red-teaming results and alignment data. The data types range from language and code to vision, voice, video, agents, simulation, robotics and the physical world.

Some tasks cross fields. A medical question asked in Spanish needs clinical judgement and native fluency; an engineering agent needs both domain knowledge and code review. We match experts to the task as it actually is.