I’ve just installed v10 and it looks amazing.
Since big companies like Google advise their employees against entering any personal or company data on any chatbot including their own, I would like to ask how are the AI auto replies processed to keep our message content safe and/if we can limit what is sent on each message by selecting only part of the text.
I’m pretty much curious how everything AI-related is processed on Em Client.
Hi! The AI BETA requests are sent directly to ChatGPT, so at this time, their privacy policy is the one that applies. In the final version of eM Client 10 we plan to make the solution similar to how our Translation works and the queries will go through our servers.
You CAN choose just part of the text to send to the AI query though - click Reply to open the usual reply window (NOT Reply with AI option!), make a selection from the original message text first and THEN click the AI button - this way only the selected text will be sent to the OpenAI servers.
This raises the question of financing, as a subscription model seems inevitable if all requests are made via em Client.
Will there be an option to store your own openAI API key?
From our point of view, the use of our own openAI API would be preferable. However, the beta does not currently allow such settings. Hence the question, is this planned?
Let me stress that AI features are great but of great concern for me.
Em Client should have big warnings regarding the usage of it and that it will send private data to a language model that is a sponge of knowledge.
Em Client should also limit/avoid sending e-mail addresses and names to chat GPT when elaborating replies with no selected text.
How courts will view cases prosecuted in relation to inaccurate claims and or losses as the result of information generated with the use of software using ChatGPT is outstanding.
However, what is clear is that there is no obvious immunity from liability where software incorporating tools such as ChatGPT is in use either by the author or software provider to the disclosure of personal information with relation to GDPR.
I just hope there is a way to turn off AI features completely. I get that it’s the new hotness, but every app is adding it, and the first thing I do is disable it if I can. EmClient is great, and the new features I’m sure will be welcomed, but please give us the feature to disable AI.
very cool. I forget that I can customize that bar! as long as it doesn’t send any metadata secretly on the backend without the user knowing, then this works.
AI is used only on user action. We don’t send any data to ours or 3rd party servers until you explicitly generate something with AI (you have to click on AI button or use Reply with AI feature).
It would be good to explicitly state it in privacy policy, as well as whether your contract with open AI has opted out using customer data for their legitimate interest purposes.
We are still considering the model for that, but as you say, these are additional (and not negligible) costs for us, so some kind of subscription for AI is very likely on the end of the day.
As “Silberfisch” asked above, there is an option needed, to use a personal openAI API key and an explicit “no AI available”.
Otherwise, I see the danger that the use of emClient in (at least German) companies could be jeopardized. The GDPR is unequivocal in this respect: no data may be passed on to third parties without consent. ChatGPT is a “third party”.
The advice from “cyberzork” to remove the function from the toolbar is insufficient: it is still available and can be reactivated without any problems.
The only thing that could be legally safe is to explicitly select the AI functionality during installation (or repair), otherwise the associated functions will be permanently deactivated.
My personal opinion:
AI support is a neat idea. However, in my opinion, it can only be operated in compliance with data protection regulations with a personally hosted AI instance.
Private users may not care, but for commercial use there must definitely be configuration options, including the explicit removal of all AI functions, according to the current state of the law.
Another question is what happens to the amount of my lifetime licenses I have bought: will they exclude AI but work with the remaining functions, or will they get worthless?
We’ll add a general On/Off switch to the app which will hide all the UI buttons/menu items once AI is switched off. We’ll also provide a way how to restrict this on a license level via License manager’s restrictions feature. We’ll also include that option loadly in the brand new initial wizard (shown when app is installed). We take privacy very seriously and we’ll be very transparent what is happening to your data and do our best to protect yout data and privacy as much as possible.