Legal implications of using public MT (confidentiality)
Thread poster: TrM Translations

TrM Translations
Hungary
Local time: 07:29
Member (2007)
English to Hungarian
+ ...
Sep 20

I am wondering if there are any adverse legal consequences of using publicly available MT solutions (Google or Bing or perhaps MyMemory, etc.).

The segments from the client documents are sent to these servers where they may get stored or even used towards improving the service at the end of the day. This suggests that confidential customer data might leak through these channels.

What is the reality of this assumption and what are the workarounds?

Istvan Fulop
TrM Translations


Direct link Reply with quote
 

Tom in London
United Kingdom
Local time: 06:29
Member (2008)
Italian to English
Not at the end of the day Sep 20

TrM Translations wrote:

I am wondering if there are any adverse legal consequences of using publicly available MT solutions (Google or Bing or perhaps MyMemory, etc.).

The segments from the client documents are sent to these servers where they may get stored or even used towards improving the service at the end of the day. This suggests that confidential customer data might leak through these channels.

What is the reality of this assumption and what are the workarounds?

Istvan Fulop
TrM Translations


I don't think they wait until the end of the day. The segments go live immediately.

A warning to all translators: don't feed the MT. If you do, you are digging your own grave - and not getting paid for doing so.

And certainly, the legal implications are considerable.

[Edited at 2017-09-20 13:41 GMT]


Direct link Reply with quote
 
The Misha
Local time: 01:29
Russian to English
+ ...
There are no workarounds Sep 20

Do not use any public MT when translating attorney work products or other sensitive legal matter. Period. Easy as that.

Direct link Reply with quote
 

CafeTran Training
Netherlands
Local time: 07:29
Member (2016)
MDÜ Sep 20

TrM Translations wrote:

I am wondering if there are any adverse legal consequences of using publicly available MT solutions (Google or Bing or perhaps MyMemory, etc.).


There's in interesting German article in MDÜ 4/2017 (http://www.bdue-fachverlag.de ), titled: Maschinelle Übersetzungs-Plug-ins in CAT-Tools, Sicher (nicht) nutzen? (sorry, I'll have to leave it to DE > EN colleagues to find a good translation for this title).

This paragraph caught my eye:

Screen Shot 2017-09-20 at 18.28.27


Direct link Reply with quote
 

Tomás Cano Binder, BA, CT  Identity Verified
Spain
Local time: 07:29
Member (2005)
English to Spanish
+ ...
Translating your question Sep 20

TrM Translations wrote:
What is the reality of this assumption and what are the workarounds?

Please allow me to give my interpretation of your question: What are the chances that I get caught sending my customer's confidential information out to the Internet and how can I avoid it?

Yes, I might have gone a bit too far, but this is what comes to mind reading your question. The workaround is simple: do not work with public MT systems and thou shalt be safe.

If you feel that MT could be a benefit to your work and your customers do not prohibit its use (as is my case with some customers), you might give a try to in-house MT, of the likes of Lilt, for instance. I have entertained the idea of giving it a try, but the inherent complexity of my work (idiomaticity, intertextuality, specialised terminology in many different areas, uncontrolled language) keeps me from trying.


Direct link Reply with quote
 

Tom in London
United Kingdom
Local time: 06:29
Member (2008)
Italian to English
Agencies Sep 21

One agency I work with has specifically instructed its translators never to use MT.

As for Lilt, isn't that a drink?

41dk9seCqNL


Direct link Reply with quote
 

Samuel Murray  Identity Verified
Netherlands
Local time: 07:29
Member (2006)
English to Afrikaans
+ ...
@Istvan Sep 21

TrM Translations wrote:
I am wondering if there are any legal consequences of using publicly available MT solutions (Google or Bing or perhaps MyMemory, etc.).


IANAL, and jurisdictions differ.

But firstly, if your client prohibits it, and you live in a jurisdiction that allows clients to make such demands of you while you retain your status as a non-employed person, and you sign the agreement, then, if you do use it, you are in violation of the agreement.

If penalties are stipulated in the agreement, you may be liable to pay them (but if this is you, then wait until a judge who has jurisdiction over you orders you to do so). If no penalties for breach are stipulated, the client may file a civil suit against you for damages suffered because of your violation, but the client has to prove actual damages (not potential damages). That said, it's easier to prove that there were damages than to prove that there were none, so don't be complacent. The client can either get a judgment against you in his own country (this has pros and cons, but mostly cons for the client), or he can hire a lawyer in your country to sue you in your own country. Either way, you'll need legal advice if a client decides to sue, and that will cost money, even if the result is knowing that you're safe.

If the agreement with your client does not actually prohibit using public MT or publicly available MT or non-local MT, you may still be in breach of the agreement if your actions breach confidentiality. And a judge will decide.

The big question then becomes: does uploading confidential information to a machine translation service breach confidentiality?

Let's be realistic: unless you send back a corrected version of the translation, no machine translation service will have *any* interest in retaining the source text that you send it. None of your confidential texts will be used in any subsequent results offered by the MT system. The only thing an MT system may want to do with your source texts would be to analyse it for keywords to calculate usage statistics. And it would be too expensive for the MT system to retain the actual text itself for any length of time.

The next question then is: do you trust the MT service? Look, it's entirely possible that an MT service out there may be a front for hackers to gather confidential texts to scan for information that they can use to make money or for political gain. But if you stick to the main ones, your only worry is whether you trust the company enough to believe their terms of service (where you can read what they do with your text).

Some public MT systems have a feature whereby your data gets an extra layer of confidentiality. Google Translate, for example, has that with their paid API.

Another type of legal consequence that has nothing to do with breach or with confidentiality is if the MT service serves a match that contains a mistranslation that you fail to correct.

The segments from the client documents are sent to these servers where they may get stored or even used towards improving the service at the end of the day. This suggests that confidential customer data might leak through these channels.


I highly doubt if these companies store your data for any length of time even if they "use" it to improve the service. The only way I can think of that they might use it to "improve the service" would be to gather statistics of usage, and that does not require storing the data for very long.

What is the reality of this assumption and what are the workarounds?


The best workaround is to use an MT service with terms of service that give greater commitment to maintaining your confidentiality. Google Translate's paid API comes to mind.

Sometimes you have to ask yourself if you trust the MT service's company's trust. MyMemory, for example, often sends your submitted text to third-party MT services, and although MyMemory's agreement with those companies is that they may not store the text, you don't know who those companies are and so you can't be 100% sure about them.

Finally, if your question was really "how can I use MT against the client's wishes and still get away with it?", then the answer is simply to make sure you never get caught. This means finding out in what ways the MT service can inadvertently give itself away (e.g. by adding non-visible characters to the translation that you don't see but your client does), making extra sure that there are no silly errors that no human would possibly make, ensuring that your translation is always of a high standard (e.g. consistent and idiomatic), and never accepting an MT match unedited unless you're very sure that you would ordinarily have translated it the same.


Direct link Reply with quote
 

TrM Translations
Hungary
Local time: 07:29
Member (2007)
English to Hungarian
+ ...
TOPIC STARTER
A bit too far Sep 22

[quote]Tomás Cano Binder, BA, CT wrote:

Please allow me to give my interpretation of your question: What are the chances that I get caught sending my customer's confidential information out to the Internet and how can I avoid it?

Yes, I might have gone a bit too far,...


Yes indeed.

What I want to understand right now is basically how to evaluate MT quality and postprocessing workflows and related issues without the burden of investing into an in-house system. A readily available approach would be to take a plugin from the SDL Store and try it out, but I don't want to do it due to the concerns I have detailed in the original question.

[Edited at 2017-09-22 13:39 GMT]


Direct link Reply with quote
 

TrM Translations
Hungary
Local time: 07:29
Member (2007)
English to Hungarian
+ ...
TOPIC STARTER
MAchine learning processes Sep 22

Thanks.

I highly doubt if these companies store your data for any length of time even if they "use" it to improve the service. The only way I can think of that they might use it to "improve the service" would be to gather statistics of usage, and that does not require storing the data for very long.


Machine learning basically incorporates the source and the human.reviewed translated segment into its system, even though just as a kind of distilled fuzzy pattern. Nevertheless I find that this is a risk of data leak.

What if you give your client an order form which clearly explains that you are going to use the MT system of company X with their terms and services available at www.X.com, do you think this is enough for a transparent process?


Direct link Reply with quote
 

Samuel Murray  Identity Verified
Netherlands
Local time: 07:29
Member (2006)
English to Afrikaans
+ ...
@TrM Sep 22

TrM Translations wrote:
What if ... do you think this is enough for a transparent process?


Why do you want a transparent process?

What if you give your client an order form which clearly explains...


Very few agency clients use the translator's order form (if indeed there is an order form at all), but instead will issue a purchase order in which their own terms and conditions are stipulated. If the purchase order's conditions contradicts those of the order form (e.g. thou shalt not use MT), and you do the job, then the purchase order's conditions apply.

Some hopeful translators issue a notice to clients to the effect that "if you offer me a job, then you agree to my terms and conditions", but in reality even if such a notice was given, it is the client's terms and conditions that will end up being used, if the client issues a purchase order with its own terms and conditions on it and you accept the purchase order by doing the job.

... that you are going to use the MT system of company X with their terms and services available at www.X.com...[/quote]

You're free to be as transparent as you wish, but what would you hope to gain?

Direct link Reply with quote
 

TrM Translations
Hungary
Local time: 07:29
Member (2007)
English to Hungarian
+ ...
TOPIC STARTER
Thanks Sep 25

Samuel Murray wrote:


Thank you for your points, I'll give them a thought.


Direct link Reply with quote
 


There is no moderator assigned specifically to this forum.
To report site rules violations or get help, please contact site staff »


Legal implications of using public MT (confidentiality)

Advanced search







Wordfast Pro
Translation Memory Software for Any Platform

Exclusive discount for ProZ.com users! Save over 13% when purchasing Wordfast Pro through ProZ.com. Wordfast is the world's #1 provider of platform-independent Translation Memory software. Consistently ranked the most user-friendly and highest value

More info »
BaccS – Business Accounting Software
Modern desktop project management for freelance translators

BaccS makes it easy for translators to manage their projects, schedule tasks, create invoices, and view highly customizable reports. User-friendly, ProZ.com integration, community-driven development – a few reasons BaccS is trusted by translators!

More info »



Forums
  • All of ProZ.com
  • Term search
  • Jobs
  • Forums
  • Multiple search