Who Is Allowed to Sign? Authorised Signatories and the Two-Person Problem
Nothing leaves an accredited facility until a nominated, authorised person releases it. In a small facility that might be two people. Here is what the rule says, what it can do to capacity, leave and succession, and how to see your own position on one page.
For: Owners and managers of small NATA-accredited laboratories and inspection bodies with one or two people who release results
Every accredited facility has a short list. It is the people who are allowed to release results.
In a large facility the list can be long. In a small one it may be two names, and one of them may own the business.
This article is about what that list can do to a small facility, and how to look at your own. It draws on NATA's published criteria as at October 2026. We have no connection with NATA, and this is not accreditation advice. Check your own scope and the current documents for your field.
What does NATA actually require of the people who release results?
NATA's ISO/IEC 17025 Standard Application Document (October 2024) puts it this way:
The facility must nominate individuals who are authorised to release results and NATA will formally acknowledge these individuals in the report on assessment.
Four things follow from NATA's criteria.
Nothing goes out without them. The same document says:
No report, whether preliminary or final, shall include results not authorised for release.
Reports can be issued electronically, including from a site other than the accredited facility, provided they "have been appropriately authorised for release".
The authority rests on knowledge, not on a title. NATA's Calibration appendix (June 2026) says personnel authorised for the analysis, authorisation and reporting of results "must have a sound knowledge of" the NATA accreditation criteria, the principles of the calibrations they perform or supervise, the relevant standards, and measurement ranges and measurement uncertainty. Other fields word this differently.
NATA wants to know when the list changes. NATA's document on the responsibilities of authorised representatives (September 2024) says the authorised representative must notify NATA in writing within fourteen days of a significant change in, among other things, the "personnel authorised to release test results".
Some fields add a named expert on top. In non-destructive testing, NATA's annex (effective August 2026) says the person responsible for technical control must be directly involved in, among other things, "authorisation of technicians as competent to perform specific tests and/or to release test results" and "monitoring of the quality of reporting of test data across the organisation". If the organisation has no Level 3 certified person in the relevant method, it "shall ensure Level 3 coverage from another source". The annex also asks for provision to delegate Level 3 responsibilities during leave.
Inspection bodies work a little differently. NATA's ISO/IEC 17020 application document (effective October 2026) says NATA will formally acknowledge the individuals nominated to release results where there is a regulatory framework or a binding agreement with a third party. It also requires each inspector to be observed on site at least once in the three-year accreditation cycle.
Is it "approved signatory" or "authorised signatory"?
Many people still say "approved signatory". It is the older term. NATA's 17025 application document refers to "previously approved signatories", and NATA's responsibilities document records that it replaced references to signatories with "personnel authorised to release test results". The NATA Rules (September 2026) still define "approved signatory", so you will see both.
In this article we say "authorised signatories" or "people authorised to release results". They mean the people on that short list.
What would we expect the rule to do to a small facility?
What follows is our reasoning, not research: what we would expect, and what a facility might find if it looked.
The rule protects the customer. Its side effects fall on the owner.
Capacity. A facility can test as fast as it has technicians and equipment. It can report only as fast as its authorised signatories can review. If work grows, we would expect the queue to form at the signature.
Leave. With two people authorised for an area, one person on holiday could halve the facility's ability to release results in that area. With one, the facility might stop reporting, or someone reviews results from a beach.
Succession. A new signatory cannot be hired ready-made. They have to show, on your methods and your kinds of work, that they can judge whether a result is valid. That is likely to take supervised time, and the supervising comes out of the existing signatory's week.
Sale. A buyer of an accredited facility is, in part, buying its scope of accreditation. If releasing results under that scope depends on two people and one of them is the seller, we would expect a careful buyer to notice.
The unwritten part. This may matter most and be noticed least. Ask a signatory what they check before they sign and you will probably get the procedure. Ask what they check that is not in the procedure and, after a pause, you may get the real answer. The reading looks wrong for that instrument type. That technician rushes on Fridays. This client's items always arrive after a long road trip. The check standard was inside limits but has been creeping for a month.
None of that is usually written anywhere. It is why the signatory is good, and it leaves with them.
An example, made up
A calibration laboratory of nine people. Two are authorised to release certificates: the owner, and a senior technician who has been there fourteen years.
The owner signs about a third of the certificates and is the only person authorised for one group of instruments. The senior technician signs the rest. Neither has taken more than a week off at a time in three years. A third person has been "coming up" for eighteen months, and has not been authorised, because nobody has had time to supervise the work that would show they are ready.
Nothing is wrong at this laboratory. Its last assessment was clean. But if either signatory left, the owner would have a hard conversation with the largest customer about turnaround.
How can you see your own position on one page?
This needs a blank sheet of paper and an hour with each signatory.
- Draw the grid. Write the names of everyone authorised to release results down the left side. Across the top, write your main areas of test, calibration or inspection. Tick each box where that person is authorised. Any column with a single tick is a single point of failure.
- Mark the horizon. Next to each name, note anything you already know is coming: planned retirement, long leave, a move to part-time. If a column's ticks all sit beside names with something coming, that column needs a succession plan.
- Count. For last month, how many reports did each person release? If one person carried most of them, write that down.
- Write the leave rule. For each signatory, write what actually happens when they are away for two weeks. If the honest answer is "they log in from holiday", write that.
- Ask for the unwritten checks. Ask each signatory for three things they check that are not in the procedure. Write them in their words. Do not tidy them. You now have the start of a document that has probably never existed.
- Write what a new signatory would have to show. For one area, list the evidence you would want before authorising someone: which methods, how many supervised reports, which unusual cases. Then estimate how long that would take if you started Monday.
- Check the notification. If the list has changed since your last assessment, check whether NATA has been told. NATA's criteria set fourteen days for notifying a significant change.
What does this have to do with AI?
Less than you might expect, and nothing that removes the signatory.
A tool cannot be authorised to release results. A person is. We could find no NATA criteria or formal guidance on using AI in the release of results. NATA's October 2026 news article on AI-supported results asks two questions that land squarely on the signatory: is a competent person reviewing the outputs, and is it clear who remains responsible for the final result? In a November 2024 news piece, NATA said assessments would need a greater focus on "organisational governance and accountability" as AI is adopted. We cover what NATA has said so far in What NATA has said about AI.
In accredited work, the person who signs answers for the result, whoever or whatever drafted it. What a tool might do is the reading before the judgement: gather what the signatory would look at and lay it out. But a tool can only be pointed at what somebody has written down. If the signatory's real checks live only in their head, there is nothing to point it at. Step 5 above is the start of writing them down, and it is worth doing whether or not you ever use AI.
We look at what an assessor would be likely to ask about such a tool in If an AI drafts it and you sign it.
One question
We are asking people in accredited facilities how this works for them. No pitch: when you are the one who authorises results, what do you check that nobody wrote down?
Write to warwick@subfrac.com. We will share what we hear, without names.
Sources
- General Accreditation Criteria: ISO/IEC 17025 Standard Application Document
- General Accreditation Criteria: ISO/IEC 17020 Standard Application Document
- Specific Accreditation Criteria: ISO/IEC 17025 Application Document, Calibration – Appendix
- Specific Accreditation Criteria: ISO/IEC 17025 Application Document, Infrastructure and Asset Integrity – Annex, Non-destructive Testing
- General Accreditation Criteria: Responsibilities of authorised representatives
- NATA Rules
- NATA assessing opportunities with AI
- Can you trust an AI-supported result?
Facts last checked:
Subfractal is not affiliated with or endorsed by NATA, and nothing here has been reviewed by it. This is general information, not accreditation or legal advice.
Who wrote this
Warwick (Woz) Heathwood is a co-founder of Subfractal.
