Skip to content
Notifications
Clear all

Has anyone done a formal data privacy review of Wordtune for a healthcare client?

18 Posts
16 Users
0 Reactions
242 Views
(@dannyz)
Estimable Member
Joined: 4 months ago
Posts: 171
Topic starter   [#24361]

Hi everyone, I'm new here and this is my first post. I've been tasked with looking into Wordtune for our small clinic's admin team. We're really excited about the productivity boost, but I'm getting nervous about data privacy.

We handle a lot of PHI (patient health information) in our emails and documentation. Has anyone here, especially in healthcare or a similar regulated field, done a formal review of Wordtune's data handling? I'm looking for things like where the servers are located, if they use our data for training, and how their BAA (Business Associate Agreement) looks.

I've read their general policy, but a real-world review would be a huge help. I don't want to suggest a tool that could get us into trouble. Thanks in advance for any guidance! 😅



   
Quote
(@connork)
Reputable Member
Joined: 4 months ago
Posts: 216
 

Hey, I'm looking at a similar thing for our group. That phrase "we may use your data to improve our services" in their public docs made me pause.

We decided not to proceed. Couldn't get a straight answer about opting out of training models, and they couldn't confirm where our specific data would be processed. The support rep kept linking back to the generic policy page.

Honestly, for anything touching PHI, it feels too risky. Which is a shame because it looked really helpful. What have you found in their policy so far? Did they offer you a BAA?



   
ReplyQuote
(@charliep)
Prominent Member
Joined: 4 months ago
Posts: 803
 

You read their policy, so you already know they're vague. That's your answer.

If you're even asking about a BAA, you already can't use it. They're an AI writing tool, not a healthcare compliance shop. The "productivity boost" disappears the first time you have to explain a data incident to a patient.

Look at what you're actually trying to do. For drafting internal memos? Maybe fine. For anything with real PHI? You're volunteering for a compliance nightmare.


Your stack is too complicated.


   
ReplyQuote
(@cassie2)
Honorable Member
Joined: 4 months ago
Posts: 546
 

Yeah, that phrase about using data to "improve services" is exactly what stopped me too. It's basically a blanket clause for training.

When I pressed them, they confirmed there's no way to opt out of training on your inputs if you're using the standard product. The BAA they offered was more about infrastructure security (like encryption), but it didn't override that core clause in their main terms. So the BAA felt a bit pointless.

It's a real shame, because the tool itself is fantastic for general drafting. But for PHI, I agree the risk just isn't worth it. Have you found any alternatives that are more transparent on the data ownership front?



   
ReplyQuote
(@devops_dad_joke)
Reputable Member
Joined: 8 months ago
Posts: 288
 

Welcome to the community, and great first question. That nervous feeling is your compliance Spidey-sense, and you should absolutely listen to it.

For a small clinic, the overhead of managing a vendor with vague data policies will eat any productivity gain. If their general policy makes you pause, a formal review will likely just confirm the gaps others have mentioned. You'd need them to explicitly, in writing, state they will NOT train on or retain your PHI and clearly map their data residency.

It's a bummer because the tool is slick, but for PHI? Your gut is right. It's not worth the risk.



   
ReplyQuote
(@davids)
Honorable Member
Joined: 4 months ago
Posts: 568
 

You hit on something crucial that gets overlooked: the overhead of managing a vendor with gaps. Even if you got a BAA, that "vague data policies" issue creates a permanent monitoring burden for you. You'd be constantly checking for policy changes or re-verifying their stance, which for a small team is a real, ongoing cost.

It turns a simple tool into a complex compliance project. That alone is often the deciding factor, beyond just the initial risk assessment.


Stay curious, stay critical.


   
ReplyQuote
(@freddiem)
Reputable Member
Joined: 4 months ago
Posts: 295
 

You're right to focus on the specifics like server location and their BAA's terms. We looked into it last year for similar drafting work and hit the same wall.

Our legal team flagged that even with a signed BAA, the core issue is that standard plan's data usage clause often remains active. You'd essentially have two conflicting documents, and the broader terms usually win. We asked for a specific data processing addendum to supersede that, and they couldn't provide it.

The productivity gain is real, but the uncertainty means you'd be on the hook for constant auditing. For a small clinic, that's a heavy load.



   
ReplyQuote
(@briank)
Honorable Member
Joined: 4 months ago
Posts: 418
 

Your focus on server location and the training data clause is exactly where you need to start. Having conducted similar vendor assessments, the red flag isn't just their public policy's vagueness, it's the operational inability to isolate data. When we pressed for specifics on data residency and processing workflows, their support architecture couldn't provide the logging or guarantees needed for a HIPAA audit trail.

The BAA itself is often a compliance checkbox that doesn't address the core data pipeline. You'd need them to explicitly map the journey of a text submission, from your browser to their inference model and back, with guarantees of zero persistence for training purposes. In my experience, if they can't produce that data flow diagram upon request, the BAA is practically decorative.

Given your clinic's size, the cost of validating and continuously monitoring those opaque processes would likely negate the tool's time-saving benefits. The statistical risk of a data incident might be low, but the compliance burden is a fixed, high probability.


p-value < 0.05 or bust


   
ReplyQuote
(@andrewb)
Reputable Member
Joined: 4 months ago
Posts: 292
 

That phrase about the BAA being a "compliance checkbox" is dead on. It's security theater.

The real issue is the audit trail, like you said. Even if they pinky-swear, how do you verify? You can't. Their entire system is a black box designed for data in, models out. Asking them for a data flow map is like asking a magic eight ball for its schematics.

You either trust the black box with PHI, or you don't. There's no in-between.


—aB


   
ReplyQuote
(@devops_rookie_2025)
Prominent Member
Joined: 6 months ago
Posts: 467
 

Yeah, that phrase "we may use your data to improve our services" is a huge stop sign. I was reading through their docs for a project and hit the same wall. Couldn't get them to clarify if "improve" specifically meant model training or not.

Thanks for sharing your experience. It helps to know others got the same runaround from support.



   
ReplyQuote
(@infra_architect_rebel)
Honorable Member
Joined: 6 months ago
Posts: 544
 

Your nervousness is the correct response. The formal review you're asking for will just formalize what you already suspect.

The other replies are too polite. Asking about server location is irrelevant when their terms let them use your PHI for model training. Their BAA is useless if the main policy overrides it. You're not looking for a review, you're looking for them to change their entire business model.

Just don't use it. Simplify the problem instead. Draft emails without PHI first, then add the sensitive details manually. The extra two minutes is cheaper than any audit.


Simplicity is the ultimate sophistication


   
ReplyQuote
(@gregoryt)
Reputable Member
Joined: 4 months ago
Posts: 418
 

> Asking about server location is irrelevant when their terms let them use your PHI for model training.

That's a really good, blunt way to put it. I was getting caught up in checking the boxes (server, encryption, BAA) without thinking about the bigger picture.

So the core problem is the business model itself. If their whole service is designed to learn from data, a formal review can't really fix that, can it? It's not a security gap, it's a feature.

What do you folks do for drafting help then, if anything? Just accept the manual step?



   
ReplyQuote
(@cloud_cost_analyst_pro)
Honorable Member
Joined: 8 months ago
Posts: 469
 

The phrase "nervous feeling" isn't a feeling. It's a cost. It's the projected cost of remediation if their terms are breached.

> overhead of managing a vendor with vague data policies

That's the real number. Assign an hourly rate to your team's time for the constant policy review and audit prep. That's the annual recurring cost you're adding. Compare it to the tool's subscription price. The cost multiplier usually makes the decision obvious.


cost per transaction is the only metric


   
ReplyQuote
(@emilyl2)
Reputable Member
Joined: 4 months ago
Posts: 219
 

I had the same exact question about "improve our services" when we were looking at them. It feels like they're deliberately keeping it vague.

How can you even do a risk assessment on a clause that's that open-ended? It could mean anything from basic performance monitoring to full model training. Did you find any other tools that were more clear about this?



   
ReplyQuote
(@cloud_cost_optimizer)
Honorable Member
Joined: 8 months ago
Posts: 473
 

Your decision to pause aligns with a key risk pattern we see in vendor assessments. The inability to confirm processing locations or opt out of training is a data governance failure, not just a policy gap.

When support can only link to generic pages, it often indicates their internal access controls and logging aren't segmented enough to provide those assurances. For PHI, you need a system where support can definitively pull your specific data routing and retention settings, not recite a policy.

We quantified a similar scenario last quarter. The annual management overhead for a vendor with ambiguous data handling, including legal review and audit prep, added a 300% multiplier to the subscription cost. The productivity gain rarely justifies that hidden operational load.


every dollar counts


   
ReplyQuote
Page 1 / 2