Skip to content

Safety

Safety and limits

A disclaimer on its own is not a safety measure. These limits are built into how the product works.

Please keep in mind: This platform provides information about clinical trials and estimates potential eligibility based on the information you provide. A match does not mean that you qualify for a clinical trial. Final eligibility is determined by the study team. This service does not provide medical diagnosis or treatment advice.

What this system will not do

  • Diagnose a medical condition.
  • Recommend a treatment or a medication.
  • Tell you to stop, start or change any treatment.
  • Guarantee that you are eligible for a clinical trial.
  • Replace your doctor, a clinical researcher, or a study coordinator.
  • Let the AI invent eligibility criteria, trial locations, contacts or medical facts.

How those limits are enforced

The model has no database access

The assistant cannot query anything. A fixed set of retrieval functions runs first, scoped to your own profile and your own results, and hands the model a finished evidence bundle. You can open that bundle on any answer and see exactly what it was given.

Objective comparisons are not done by AI

Age, sex, condition, stage, prior treatment and biomarker checks run through a deterministic rule engine in plain PHP. The same inputs always produce the same verdict, and it can be re-run and audited.

Every structured criterion keeps its source text

A parsed criterion that cannot point back at the sentence it came from is not stored. When the LLM parser is enabled, anything it returns whose quoted source does not match the input statement is discarded.

Generated text is checked before you see it

Explanations and answers are scanned for eligibility claims. Any output saying you are eligible, that you qualify, or that anything is guaranteed is rejected and replaced with the deterministic version.

Out-of-scope questions are intercepted early

Requests for diagnosis, prescriptions or treatment changes are caught before any model is called, and answered with a redirect to your doctor or the study team.

Similarity is never evidence of eligibility

Semantic matching decides what to show you and in what order. It contributes to the score, but it can never on its own satisfy a criterion.

What a match status actually means

Potentially Eligible

A strong match with no known major exclusions, and no unresolved criteria we could have checked. It still is not a decision about you.

Possible - Missing Info

Promising, but at least one criterion we could have evaluated is unknown because we do not have that information from you.

Likely Not Eligible

A known major exclusion criterion appears to be met. Study teams do sometimes make exceptions, so this is a signal, not a verdict.

If you are in urgent need

This service is not for emergencies and no one monitors it in real time. If your symptoms are worsening or you need medical help now, contact your doctor or your local emergency number.