mpathic AI

mpathic AI

A clinician-led AI safety evaluation and benchmarking platform

Officially listed
English
Paid
AI基准评测
0favorites
3views
Promote this tool
mpathic AI screenshot

Tool details

mpathic turns clinical judgement itself into a product: more than seven thousand licensed clinicians score models' high-risk conversations, and software scales that judgement. Its customers are frontier model labs, enterprises and health-tech companies.

Core Capabilities

  • Clinician-led evaluation: test scenarios are written by licensed clinicians, with both explicit and subtle risk signals across multi-turn dialogue, covering suicide, eating disorders and misinformation; two thirds of the expert network hold doctoral-level clinical credentials and half still see patients.
  • mPACT benchmarks: an in-house framework with three public benchmarks that score detection, interpretation and response, with severity weighting for the riskiest scenarios.
  • No AI judges: every score comes from trained clinicians rather than models grading models, with inter-rater reliability calibrated above 0.80, plus drift tracking and cross-demographic bias audits.
  • Red teaming and policy design: taxonomy, rubric and policy writing that turn "what counts as harm" into annotatable rules.
  • Quality platform: conversation quality as measurable metrics, including live session insight, PII redaction, speech-to-text and API integration that plugs into your own systems.

Use Cases Before launching a companion, mental-health or health-advice product, run the suicide, self-harm and eating-disorder scenarios through it; it also answers regulators and investors asking whether third-party evaluation was done. Life-science teams use it to turn interview and consultation quality into trackable metrics.

What Sets It Apart Most evaluation platforms let models grade models; mpathic sells clinicians' judgement plus a behavioural-science framework. Its scoring is deliberately clinical: missing a risk matters more than over-flagging, which is the opposite of generic safety leaderboards. What you get is reproducible clinical-grade evidence, not another scoreboard.

Editor's Take This is one of the few evaluation services that treats people as the core asset rather than a bottleneck, and it suits teams that take psychosocial risk seriously. The cost is explicit: no public price list, no self-serve signup, everything starts with a call, so small teams can hardly test the water. Its impact numbers are mostly customer-case claims without auditable third-party evidence.

Pricing

Custom pricingCustom (quote on request)
Expert evaluation project
Custom (quote on request)

Quoted by model scale and number of evaluation scenarios

mpathic Platform
Custom (quote on request)

API and dashboards, provisioned per project

Additional notes

No public price list and no self-serve signup; the platform states compliance with SOC 2 Type 2, HIPAA, GDPR and 21 CFR Part 11.

FAQ

No. There is no public price list and no self-serve signup. Every engagement is quoted per project after a discovery call, priced by model scale and the number of evaluation scenarios.

User reviews

User Reviews

No reviews yet

No one has reviewed this tool yet