ZynoviqWhat we do not do.
Most suppliers tell you what they can do. This page tells you what we will not, because in professional work the limits matter more than the claims.
We certify our answers, not your system.
We provide
- Professional ground truth
- Evaluation findings
- Regression evidence
- Red-team findings
- Improvement data
We do not provide
- Certification, approval or assurance of an AI system
- Regulatory or statutory certification
- Statutory audit
- Legal or tax opinion on your affairs
- Cybersecurity or penetration-test certification
- AI-safety certification
Every answer in our benchmarks is certified: solved independently by qualified professionals, reviewed, adjudicated where needed, and tied to the law with a valid-as-of date. That is what we stand behind. What you get from us is evidence of testing, honestly labelled. If your regulator or board needs independent assurance, it should come from an organisation with no stake in the result. Evidence.
We do not build, run or fix your AI
This business owns no GPUs, trains no models, hosts no models and deploys no agents. We have no platform to lock you into and no reason to prefer one kind of AI over another. We report what is wrong and what correct looks like; your team makes the changes. The AI boundary.
Zynoviq's separate Enterprise AI division does build AI products. We never test, grade or produce data for its products, and our delivery teams are separate. How we stay separate.
We do not touch production
Tests run in your sandbox or test environment, with scoped, time-bound, revocable access. We do not connect to production systems and we do not need your production data. Every test case is synthetic; never a real taxpayer, company, employee, patient or customer record, yours or anyone else's. How tests run in your environment.
We do not open another country's law until the professionals are in place
Our ground truth rests on professionals who hold the relevant credential and have demonstrated capability in the specific area. We will cover UAE VAT, Singapore GST, IFRS or ACCA markets only where qualified professionals in those regimes are qualified and calibrated. Until then, they are labelled Planned. Domains and their status.
Three more limits
- We do not supply people. We deliver defined work with defined outputs, not staff by the day. The expert review queue covers flagged decisions, not your whole volume. Expert review queue.
- We do not measure our own data and call it proof. After your team uses our data, we can re-test and report what moved. That is a before-and-after measurement. Training and test separation.
- We do not publish your results. Results on a customer's system are never shared or published. Our public research covers only public AI models. Research.
What we do not do: questions people ask
Why do you not certify AI systems?+
Because we also sell data that improves AI, and because Zynoviq builds AI products in a separate division. We are not a disinterested party to whether your system is good enough, and we will not pretend to be. We certify our answers and give you honest findings instead. How we stay separate.
Who can give independent assurance on our AI?+
An organisation with no stake in the result, such as an audit or assurance firm. Our evidence record is written so they can rely on it without us. Evidence.
Will you fix the failures you find?+
No. We tell you precisely what is wrong, why, and what correct looks like, and your team makes the changes. Where the fix needs material that does not exist, we can write it as data for your team to use. Improvement data.
Can you test strategy or commercial judgment?+
No. Where qualified professionals would reasonably disagree, there is no correct answer to mark against. Genuinely ambiguous cases are reported separately and never scored as failures.