Sorry if this has been answered before. I read the BeMyEyes privacy policy and I'm wondering if I understood it correctly. As I understand it, if you use Be My AI, the images and text you send are used to train the underlying models and you can't opt out of that, is that correct?
Comments
Deleting images
Hi,
If you live in the EU or the UK, you can ask them to delete your images if you don't want them stored. That's because of the data protection laws here. I don't know what happens if you live outside these juristictions. I don't think you can get all your images automatically deleted from the server every time you take a picture. Recently I asked them to delete a picture, but they actually ended up deleting my whole image history with them which was fine. It took them a few days to get back to me.
Tara
Thank you! I live in the EU so that is no problem, however, I feel that a option to opt out of training (or even better, an option to not retain pictures at all outside the current conversation) would be great. As it is, I'm a bit hesitant to use Be My AI as much as I'd like because I might inadvertently include sensitive info about me or someone else that then will become a part of the model (it has been shown that it is possible to access such sensitive info if you prompt the models in the right way). Other services, like Seeing AI and PictureSmart in JAWS, does not save your images or train their AI models with them as far as I can tell.
So very much this
I would love to use AI for describing certain things, like artist made images, and do not do this most of the time because the of the way OpenAI indescriminately uses and scrapes for data, because i know many artists do not want their work contributing namelessly to the model's success. Having an option not tod o this in any AI solutions would be a huge sstep forward.
API data isn't used for training
The OpenAI Ts&Cs are clear that data uploaded via their business tools including the API is not used for training. Since Be My Eyes is using the ChatGPT models via the API I would therefore expect data sent to it to be covered by these terms.
Relevant part of the privacy policy
Here is the relevant part of Be My Eyes' privacy policy:
Artificial Intelligence
Be My Eyes offers blind and low-vision users the ability to access third-party artificial intelligence (“AI”) tools to analyze and describe images and videos that they submit. A third party provides the AI engine powering our AI Service. If you use our AI Service, the images and videos you submit, including any personal information they contain, will be transferred to, processed by, and may be stored by that third party. We will use the information you submit through our AI Service to provide visual assistance Services, including descriptions of the images or videos. For example, if you submit a picture of the inside of your refrigerator, our AI Service can help identify what’s in it. We may also use the information you submit as described under the heading Video Streams and Images, above. Please note the third-party AI provider may also use the images and videos, including any personal information they contain, to help train its AI engine (for example, to recognize more images).
Here is the link to the policy:
https://www.bemyeyes.com/privacy
Follow-up
After a bit of digging I found the Be My AI FAQ on the Be My Eyes website. In this FAQ it states that OpenAI does not use the data to train their models, which is great. However, because of that, the FAQ and the privacy policy contradict each other. Another thing from the FAQ is that Be My Eyes, during the Be My AI beta period, stores images for review etc., and it is not clear to me if Be My AI still is in beta and therefore this still is true. I have sent all these questions, including my original question of whether you can opt out of them storing images, directly to Be My Eyes and I hope they'll clarify things.
The Privacy Policy vs the FAQ
It could be the case that they just wish to reserve the right to let the third-party AI tool store and process the images and data sent, while it might still be true that GPT/OpenAI does not do that for the time being. Be My Eyes might even switch to some other AI model or offer multiple options sometime in the future.
Re: Enis
Yes, I think you're right, but I want that to be clarified as it isn't crystal clear right now.
Emailed Be My Eyes
I emailed Be My Eyes about this a week ago but no answer yet.
photos
You mean that the pictures I ask be my eyes are not deleted after I got a description but they are some where else?
hey, am so glad you emailed…
hey,
am so glad you emailed. live in a place where data protection laws aren't that good. and wouldn't want my images to be stored. i would also love to have the option to opt out. since open AI and things like chatgpt are being used. till then, ig i'll use it a little less. i use it less but friends who regularly need help use it a lot more so maybe we'll just avoid using it for personal or sensitive images till then.
As I understand things
The images you send to Be My Eyes are *not* used to train OpenAI's models, however, they are stored at Be My Eyes for "review" during the beta period. I don't know if the beta period is still going on.
Below I post the relevant part of the FAQ where this is mentioned (look in one of my earlier posts for the part of the privacy policy that concerns this):
How are photos and chat interactions processed and stored, and who will have access to my data?
Be My Eyes initially processes your photos and chats to (a) send them to OpenAI for further processing, (b) provide Be My AI’s descriptions, and (c) give you follow up options. OpenAI processes the photos and chats to provide the AI response.
During our beta testing, Be My Eyes stores all photos and chat interactions on encrypted servers so that we can review them, if necessary, based on tester feedback. That helps us build a better service based on your feedback. It also allows us to address any safety or content issues that may come up. OpenAI does not store the photos or chats after processing them. None of your photos or chats are being used to train the AI.
The storage and use of your photos and chats by Be My Eyes are governed by our privacy policy.Â
that's what. this is just…
that's what.
this is just vague. and more vague for those of us who don't live at places having good data protection laws.
don't know what the storage period is.
The iOS shortcut
If you would like to bypass be my eyes and skip directly to open AI, there is a shortcut somewhere, I believe it’s on the main page right now, this will allow you to jump directly to using the API, though it will cost you.
thanks so much
thanks so much
Interacting with these LLMs without a middleman is usually best
Be My AI was amazing when it was the only way to access GPT-4 Vision, but now that it's publicly available for all to use, having a direct connection with the AI provider is better not only for privacy reasons but because you can generally get better responses out of it, since you can better teach it your needs and expectations.
Tools like VOCR, AI Content Describer for NVDA on Windows, or my shortcuts are great ways to access these AI models directly without needing to know how to code. The hardest thing about the process is setting up a billing agreement with OpenAI but even that is really easy. You sign into your account (you can sign in with Google If you don't want to make an account directly,) drop $5 into the account and you're basically set for a couple months, probably longer now that we have GPT-4O.
OpenAI doesn't train on images you send to them through the API, which is what you're doing when you use one of these tools. There is no middleman here, either. The image goes straight from you directly to OpenAI.
Don't get me wrong, I think Be My AI is amazing, especially because it's free, but if you can spare $5 every now and then, I think you'll be much happier using one of these other tools, even if it's only for a few images here and there.
AI
Do the 2 AI in JAWS picture smart send the photo to be check? Both do a great job and now that you can ask question is great.
Answer from Be My Eyes
I got an answer. Not the one that I wanted but still. I paste it here, along with my answer to them:
"Thank you for your questions. We do not use your images to do any training as it is right now, but we may do that in the future as our terms say. Be My Eyes is in beta and we do store your images.
There is no way to opt out of storing your images. We can delete your account if you want to, and that will also delete the images."
And here is my answer:
"Thank you for your answer. For what purpose and on what legal grounds, in terms of e.g. GDPR, are the images stored?
I will not terminate my account for now, I am however considering it due to this, to me, a bit shady privacy practice. The images that are uploaded may, inadvertently, contain sensitive information since many of us that use this service are blind, we cannot always control what we’re taking pictures of."
This makes me a bit uncomfortable using Be My Eyes I must say. I don't understand why they must store all images that we send to them, especially indefinitely, and why we can't opt out of that. When we've gotten our description of an image, just delete it. If people want to help Be My Eyes by sharing their images and conversations, absolutely let them do so, but do not force this on people please.
I compare this to SeeingAI which also says that their "Scene" feature, which for all intents and purposes is the same as Be My AI, is in "preview", but the info in the app clearly states that the uploaded images are removed when the conversation ends.
They clearly need to work on this
I hope the company takes these concerns seriously. Else I am not using Be My AI for descriptions.
Not really looking for alternatives
For me, the important part is that we, as a community, knows what the companies we are interacting with, especially those who we trust with potentially sensitive information, do with our information. As I said in my earlier post and have stated before, we have a problem sighted people don't, we can't always tell if we take images where sensitive information can be seen because we don't know if it's there, therefore, in my opinion, companies that cater to us specifically should take extra care about privacy when it comes to data we send them. Just saying that "you should not provide sensitive information" is not good enough I feel.
There are alternatives, both free and paid, so I'll find ways to get good image descriptions. AIRA is not available in Europe as far as I know.
Nonsense!
I have one and only one simple, short, clear, straightforward question: How on earth am I supposed to know what is featured in an image in which I have no clues as to what is shown and would like to have described by Be My AI or whatever app/AI model? Is this already not exactly why I want and need a description of the image? Uh, okay, I know I asked two questions and neither were short or straightforward. By the way, it is also for the very same reason that we should not be held liable for sharing personal data when uploading images to get their descriptions. It is already after we get those descriptions that we find out what the uploaded photos contain. Besides, would it not be a violation of our privacy to have to ask someone sighted to describe what (s)he sees in a photo? In conclusion, the issue will most likely remain unresolved unless or until we all have access to accurate and quick on-device descriptions without having to upload the images to any remote servers.
concerning response
I don't like the sound of their response at all.
This is really concerning because we know OpenAI isn't doing this.
OpenAI Enterprise Policy says:
If the AI platform providing the service isn't doing anything with our images, why on earth does Be My Eyes feel like they have to?
Enes and Aaron
Enes: Exactly! An important thing is also, it is not only our own personal information, it might be others' as well. As you say, if we get on-device descriptions (which seems very far away considering the processing power needed), then the problem will be solved, but even if we must send images to a remote server, the companies *can* solve it by implementing good policies, if they want to.
Aaron: I agree, I don't like the tone either. I hope they'll come back with some explanation and the legal grounds for this. Since GDPR in the EU is rather strict on what info you are allowed to keep it will be interesting to hear what they have to say.
Speaking of OpenAI, I asked them a while back how they handle images that are uploaded in ChatGPT (the non-business version). It took them a while to respond, but their answer is that, contrary to their own documentation, files (including images) are retained only as long as the conversation they're linked to, if you have opted out of model training of course. So, because *everything* you send to OpenAI is kept for 30 days after deletion (including via the API), for security/monitoring purposes (which I'm totally fine with, they must protect their systems after all), the maximum time your personal info will be there is 30 days. In their documentation it is stated that images are kept "indefinitely", but according to the mail I got that is not true.
this is very concerning…
this is very concerning.
that's why am pretty sceptical about AI coming more and more into the market. esp in places where we still don't have clear and good IT laws.
Can you say selling data!
Would not be surprise if the data is not sold to others for ads.
They might not train a model…
They might not train a model themselves, but they could transfer images to other research organizations for RnD.
I have no idea what they mean by to develop products and services that may assist other members of the general public.
Here are some direct quotes below:
We record and store video streams and other images (i) to enforce our Terms of Service, (ii) to promote and preserve safety, and (iii) to improve our Services and create new Services.
We may provide recorded video streams or images to other organizations that are performing research or working to develop products and services that may assist blind and low-vision people or other members of the general public.
If you use Be My AI, the images you submit will be processed by our third party artificial intelligence provider.
If your video or image contains personal information - for example you say your name on the video or show mail with your home address - that information will be included in videos or images that we store and use as described above.
this is yikes
this is yikes
Nothing is really private in today’s world
I really don’t care. You probably have more sensitive information about you on social media platforms and internet than Be My Eyes. Not to mention that you are using smart phones and if someone really wants then they can have access to all of your smart phone data, files, pictures etc. Deleting things won’t help either because they are only deleted for ourselves but they are not really gone.
Right, but...
The thing is, when a company accesses, collects, stores, processes, shares and/or sells our personal information, it can base that on its vague privacy policy and practices and assert that it is us who consented to that, making it easier to justify its policies and more difficult for us to object to them; whereas someone doing those things without our express consent will likely face more legal issues and sanctions, and we will at least be able to claim that we never consented to any of their unlawful acts.
There's one phrase that makes me so concerned and annoyed: "including but not limited to". I encounter it in most such agreements where they desperately look for ways to make things as ambiguous and open to interpretation as possible.
Mr. Mosen
For anyone who subscribe to livingblindfully can you send an email to him about this question? He suppose to have good relationship with those people in charge of Be my eyes. This way this question is answer one way or another. I am not a subscriber to it.
How are photos and chat…
How are photos and chat interactions processed and stored, and who will have access to my data?
Be My Eyes initially processes your photos and chats to (a) send them to OpenAI for further processing, (b) provide Be My AI’s descriptions, and (c) give you follow up options. OpenAI processes the photos and chats to provide the AI response.
During our beta testing, Be My Eyes stores all photos and chat interactions on encrypted servers so that we can review them, if necessary, based on tester feedback. That helps us build a better service based on your feedback. It also allows us to address any safety or content issues that may come up. OpenAI does not store the photos or chats after processing them. None of your photos or chats are being used to train the AI.
The storage and use of your photos and chats by Be My Eyes are governed by our privacy policy.
So I don’t see any problems there and the calls with volunteers are recorded anyway for the safety reasons.
Lottie
Can I have it! Needed so my cat can not send messages to my brain or trying to control me.
This passage is the problem for me
"During our beta testing, Be My Eyes stores all photos and chat interactions on encrypted servers so that we can review them, if necessary, based on tester feedback. That helps us build a better service based on your feedback. It also allows us to address any safety or content issues that may come up." It is not stated anywhere for how long they will be stored (e.g. when will Be My AI go out of beta and will the policy change then?) and who have access to the images and conversations. "Encrypted servers" is also really funny in this context seeing as most servers today are encrypted, that is not something that should have to be stated.
I don't mind at all that people don't care about their privacy (although people around them might, remember, it is not only your own privacy but the people you interact with as well), that is their choice. Some will see me as paranoid and that's fine too.
Holger: I am no subscriber either but it would be interesting to hear him ask Be My Eyes about this. They might have a good explanation, and if so I'll rest my case, but as long as their policies and answers are vague and lack some important information I'll be a bit vary.
It is not in beta anymore
Be My AI is not in beta anymore. Even the beta button next to the Be My AI button is removed and everyone can access this feature even new users so it is not in beta anymore.
A different perspective
Hi all,
While I do agree that people should be able to choose whether they want their images used for training purposes or not, there's a take here I don't really agree with. Enes you say, 'we should not be held liable for sharing personal data when uploading images to get their descriptions. It is already after we get those descriptions that we find out what the uploaded photos contain.'. I would say it depends on what you already know about your surroundings. When I take a picture of food packaging or something I want identified, I take said picture in an area where I know I haven't got any paper documents lying about with identifying details on them. Unless you're trying to identify documents, take a selfie or a picture of someone else, what could you possibly be taking a picture of where you wouldn't know whether it could potentially hold identifyable information or not? The only exceptionns to this rule are if you're living with other people and they could have left personal stuff lying about you don't know about, or you're out in public and you want to take a picture of something. In these two scenarios, you should absolutely be aware that personal stuff might get picked up, and if you take a picture outside, random people might end up in the background who wouldn't want to be there. Everybody knows this, right? So to an extent, we should be held liable. We know our surroundings, and we know that when we receive paper documents through the post, they'll contain personal details. It's always recommended by Be My Eyes not to knowingly take pictures of personal documents anyway. If you're really not sure what something is, use Seeing AI. But I definitely think people should have the option to choose whether they want their images stored for training purposes or not. I can opt out of using my images for training purposes in the ChatGPT app. I don't usually care what images Be My Eyes stores apart from one time I took a photo with personal details. I knew this before taking it, but I needed a good description. I knew I could email them and get it deleted anyway.
I agree with Tara
Hi Tara
You have really good points and I agree with you.
About public places, random people can appear everywhere. For example, sighted people are posting stuff on internet anyway that have random people in the background. It is not uncommon. Plus if you do have people in the background then they usually are not in full view. Also, many public places have cameras so I think that we all have appeared somewhere whether we want it or not.
It is basically the same like we all know that our houses and gardens are seen on Google maps and after a certain time the governments are taking new pictures about the landscapes and updating them on Google maps. Well, maybe someone doesn’t want that everyone can see their family’s car, their house and things in their garden but it is a reality and not for us to decide.
My point is just that we are not snowflakes who are living in a cotton. Unfortunately, this world is not as private as we would like it to be and it won’t never happen.
Tara
I agree that we have a responsibility to minimize the amount of sensitive info we send and we can, to some extent, be held liable if we send such info, however, we have a harder time identifying such things than sighted people have and therefore I feel that companies that make services for us should take more steps to ensure privacy. After all, as with SeeingAI, it can be done, so why not do it? I also agree that a real simple and good solution would be to give us the option to opt-out, if they do that I'll be really happy, because Be My AI is a great service, fast, detailed and free. A question, have you asked them to delete your images and did they do it? In my reply I got the impression that I had to delete my account for that to be possible.
Tarja: According to the mail I got (look at on of my posts above) it is in beta still. It is not indicated anywhere so I am a bit confused about that, but that is what Be My Eyes themselves say. Also, I don't agree with your viewpoint, just because some companies treat privacy lightly (including many of the really big tech companies) is not a reason for more companies to do it.
Seeing AI seems different
Well, Seeing AI seems different. Its AI doesn’t have so wide knowledge as Be My AI. Seeing AI is good for documents and short text. Also, I quite don’t believe that Seeing AI won’t store the information because you have a feedback button there where you can react if you liked or disliked the description. Why is it there then? Also, we can’t forget that in Be My Eyes it was the blind community who requested the link sharing of the images and descriptions. I really like the sharing option but in Seeing AI you can’t share anything unless you save the pictures and texts separately.
re: blindpk
Yes, Be My Eyes did delete my images after I emailed them and asked them. I asked for one particular image to be deleted. I gave them the day and the approximate timeframe I took the image. They deleted all the images in my account though. My account wasn't deleted. I'm still using the account I've been using for the last few years no problems. It's surprising they're telling you something different actually. I don't see why they'd need to delete your entire account so your images can be deleted. That certainly didn't happen for me. My account is still active and working fine.
Response I got from the AI
I took a random picture then asked it what be my eyes does with the photos and got
Me: What does this service do with the photos i take? Are they stored anywhere?
Be my ai: The photos you take are used to provide you with descriptions and assistance. They are not stored permanently and are only used for the duration of the session to help answer your questions. If you have any concerns about privacy or data handling, you might want to press the "Call a volunteer" button for more detailed guidance.
Me: Are they stored on my account or on be my eyes anywhere?
Be my ai: The photos you take are used temporarily to provide assistance and are not stored permanently on your account or within the Be My Eyes service. They are deleted after the session ends. If you have specific concerns about data storage and privacy, I recommend pressing the "Call a volunteer" button for more detailed information and reassurance.
So thats some positive information.
I hate to break it to you,…
I hate to break it to you, but if that's not in the terms of service in the Be My Eyes app itself, it is not accurate. AI is prone to making things up and you cannot rely on it for legal things like this or to tell you the truth about something like a terms of service agreement. What it says in the Be My Eyes TOS is what they will be doing with your data, not what the AI says at this moment.
Re: Tara
Very interesting. I haven't heard back from them yet about this, so I emailed them now asking if their policy has changed since you were able to get your images deleted. I also repeated the question about the legal basis of them storing our images indefinitely. I hope they answer soon, I feel that their support takes a long time to get answers from in general (have had contact with them before and it has always been a bit slow, maybe they don't have enough support staff).
No, asking the AI about their policies does not get you anywhere, most likely OpenAI does not have access to the Be My Eyes policies anyway, so it wouldn't know (even if it had it wouldn't be a safe bet, in these situations, a answer from the company directly is needed).
Honestly stuff like this…
Honestly stuff like this makes me much less likely to use a product at all. This kind of behavior, especially being dircted at a more vulnerable group, feels like we're being preyed upon and taken advantage of. I still have Be My Eyes for when I need to call a volunteer but I will not touch their AI feature with a ten foot pole after reading that we have no option to opt out of having what may be very sensitive information about us on a database (like prescription labels or documents we may not know are visible in a pic). If I could afford Aira I would ditch Be My Eyes completely for being so shady about what they're doing with our information and not giving us, the people who are the whole reason Be My Eyes exists at all, just a group to harvest data from for no good reason they will say.
Re: Magic Retina
I feel the same way and I don't understand why they do this. I can add that they still haven't got back to me about this.
As far as I can tell, right…
As far as I can tell, right now companies are doing it because AI has been really hyped up and they're trying to ride a possible wave. But also there are no firm laws or regulations stopping them from doing things that are honestly very morally awful, like literally stealing personal private information from the blind that may put those customers at risk. When companies aren't stopped from doing things that are harmful they will keep doing them whether they can see a direct benefit at the time or not, history has proved that repeatedly. Hopefully the law will catch up sooner rather than later on stuff like this and there is every sign that the AI bubble is about to pop, though how much damage it will do when it does remains to be seen. For now, if something doesn't feel right to you, opt out of that service and don't be afraid to tell the companies why if they ask.
If you wouldn't ask a stranger on the street to read your mail to you, don't ask Be My AI to do it. At least with your neighbor that person won't remember every detail and potentially sell it to a data broker company.
An update
Yesterday a mail was sent out from Be My Eyes regarding changes to their privacy policy and some other things. I'll summarize it here:
1. They want to use video conversations to train future AI models to make them less biased towards blind people. You can opt out of this. I think the idea is interesting but for me my privacy is more important.
2. They clarify their privacy policy a bit regarding e.g. that photo data is not used to train AI models (which was not totally clear before). They also write that from January 2025 they will set a limit on how long they will save users' photos and conversations.
For me personally, I appreciate that they seem to take privacy a bit more seriously and strive to make it clearer what they do and do not do, however my main concern is still not addressed, namely, they will still save photos and conversations. The change that will come in January 2025 about limiting the time that they store the data is welcome, but I still have a hard time seeing why they should store them in the first place. They say it is for "improving their services" and so on, but the Be My AI feature is nothing more than an interface to a underlying AI model, and if the AI company does not save the photos, which they don't, why should Be My Eyes do it? So, while they are taking steps in the right direction, they don't go as far as I'd hoped.
Why saving photos?
My concern would be why do they need to save them? If we wanted to be save that would be different, if the app pops with a message do you want to save the data? yes or no.
I was also told the same stuff.
The reply I got also stated that they wouldn't use my data for model training and that I could request my conversation history to be deleted. When I asked why I had to contact them instead of just being able to do that myself within the app, I was told that they did plan on adding such an option.
Update
I have still not gotten any clarifications from Be My Eyes regarding this (I have sent a couple reminders, the last one a few weeks ago). I have not seen any option to remove my conversation history in the app yet and, as I said before, the updates outlined in the mail that came during the summer are not addressing the main question, why are photos and conversations stored in the first place when the AI model doesn't? What more specifically are they used for?
Here is their official blog post
https://www.bemyeyes.com/blog/be-my-eyes-launches-initiative-to-make-ai-models-more-inclusive