Hello Google HAI-DEF team,
Thank you for releasing MedASR , it’s a fantastic model and has been very valuable for our work.
We recently built and released an iOS offline-first AI scribe called Medavo, which uses MedASR as part of its speech-to-text pipeline:
apps.apple.com /ie/app/medavo/id6764593474
Early feedback from clinicians has been very positive, particularly around medical transcription accuracy in real clinical conversations.
Thanks again for the work on MedASR.
Regards,
Mary
Hi @Mary_Colm -
Thanks for sharing this! Glad to hear that MedASR is being used in a real clinical workflow, and congratulations on the Medavo launch. We appreciate the feedback from clinicians and are glad to see the model performing well in practice.
Congratulations! @Mary_Colm and team! We recently launched the HAI-DEF Showcase. I’d like to encourage you to submit your use case . Here is our soft launch announcement.
Thank you for building with us!
Thank you for the kind words and for making MedASR available to the developer community. We’re excited to be using it in Medavo and look forward to continuing to build on it.
Thanks so much for the encouragement, really appreciate it! We’ll definitely submit Medavo to the HAI-DEF Showcase and look forward to sharing our use case. Excited to keep building with MedASR.
Why is Google publicly promoting and encouraging the use of MedASR in “real clinical workflows” when Google’s own Terms of Service explicitly prohibit this?
The Gemini API Additional Terms of Service clearly state:
“You may not use the Services in clinical practice, for medical advice, or in any manner that requires regulatory approval for medical devices.”
Additionally, the Health AI Developer Foundations (HAI‑DEF) documentation explicitly warns:
“MedGemma is NOT an approved medical device. It must not be used as the sole basis for clinical decisions.”
So how can Google engineers congratulate developers and even invite them to a Showcase when their software uses MedASR in exactly the kind of environment Google itself says is NOT a medical device?
Are you endorsing violations of your own Terms of Service?
Or are you simply ignoring the regulatory requirements (EU MDR, HIPAA) that govern the use of AI in clinical environments?
Using a non‑certified model for “real clinical conversations” is not a “fantastic use case.”
It is a patient safety risk.
Explain why Google’s engineering team is promoting the use of your models in ways that your own legal and compliance documentation strictly forbids.
You are directly violating the Google Health AI Developer Foundations Prohibited Use Policy.
I am looking at your official policy right here:
https://developers.google.com/health-ai-developer-foundations/prohibited-use-policy
It states clearly: ‘You must not use the Services… in any manner that is overseen by or requires clearance or approval from a medical device regulatory agency.’
Fereshteh Mahvar (@fmahvar), you just congratulated a team for using MedASR in ‘real clinical workflows’ (Medavo app). By doing so, you are:
Explain why a Google representative is promoting the violation of Google’s own health AI safety rules. If this is not a ‘manufacturer’ risk, then what is? You are actively exposing patient health and your own company to catastrophic liability.
The Medavo application advertises itself as an “Offline Medical AI Scribe” and is designed to transcribe medical conversations. This functionality qualifies as a medical device in every jurisdiction — EU, USA, Middle East, Asia — because it produces documentation that influences clinical decision‑making. The definition of a medical device is global and not region‑dependent: IMDRF, WHO, EU MDR, and the FDA all classify software that generates clinical documentation as a medical device.
Medavo holds no CE certification, no FDA 510(k) clearance, and no internationally recognized medical device authorization. Therefore, the product is considered an “unauthorized medical device” in every country.
The Apple App Store is globally responsible for medical device compliance, regardless of whether the user is located in the EU, the USA, or Dubai. Platform responsibility is not region‑limited.
Google’s own Health AI Developer Foundations documentation explicitly states that MedGemma/MedASR is “NOT an approved medical device,” and that clinical use of Google models is prohibited worldwide. This prohibition is global and not restricted to the EU or the USA.
Despite this, two Google employees (posting with Google badges and official Google profiles) openly congratulated the use of MedASR in a real clinical workflow and the clinical launch of the Medavo application. This is not user‑generated content — it is Google‑generated content that directly endorses the clinical use of a non‑certified medical device.
This constitutes a violation of Google’s own global policy and amounts to platform‑level clinical endorsement, regardless of whether the user is in Dubai or anywhere else. Public validation of clinical workflows by Google employees creates worldwide legal exposure.
If Medavo makes an error (e.g., hallucinating a drug name or symptom) and a patient is harmed, the liability chain is clear:
Clinicpoints Ltd. — the manufacturer of the software
Apple — the platform that allowed distribution of a non‑certified medical AI
Google — the platform whose official employees publicly endorsed its clinical use
This is not region‑dependent, not Dubai‑dependent, and not location‑dependent. It is a global medical device compliance issue.
It is provably NOT the user (Mary_Colm) who posted the HAI‑DEF link in the thread — it was the Google employee (fmahvar). This means Google invited Medavo into the clinical HAI‑DEF Showcase. This is evidence that Google is supporting clinical use. Google’s own policy explicitly forbids clinical use: “MedGemma is NOT an approved medical device.” “You may NOT use the Services in clinical practice.”
Let’s summarize the absolute rot at the core of Google’s AI ecosystem right now:
So, let’s connect the dots:
Google’s own policy says ‘do not risk patient safety’. Google’s engineers say ‘congratulations on using it in a live clinic!’. And the developers openly admit ‘we don’t give a damn about consequences, we just want our money’.
This is a complete regulatory collapse. You are enabling and encouraging uncertified, non-compliant medical AI tools to be rushed into real clinics, while everyone involved washes their hands of any patient safety liability.
It’s sickening, it’s hypocritical, and it is a massive disaster waiting to happen. The paper trail is locked, the archives are permanent, and the whole world can see what you are doing.
Here are their posts on LinkedIn:
https://archive.ph/AwSzq/image
