Photos carry information text never does. A photo of a tax document shows your employer, your income, and your face. A photo of a screenshot leaks the URL bar and the open tab. A photo of a whiteboard carries handwriting that OCR turns into searchable text. A photo of your kid carries their face, their school uniform, and the location metadata embedded by your phone.
That is why the question “can I upload an image to an AI chat” is its own privacy question, separate from “can I send text”. Every vision-capable chat handles the upload, the temporary processing, the retention, and the training pipeline differently, and almost none of them publish a setting that is specifically about images.
This page is a privacy-first comparison of the hosted AI chats that actually accept image uploads today, plus the local fallback that beats them on all of the above.
TL;DR
- Image uploads are a different threat surface from text. Photos carry EXIF metadata (GPS, device, timestamps), embedded text that OCR can lift, and the biometric identifiers in faces and handwriting. Choosing a chat on the basis of its text-chat policy alone misses the upload itself.
- Anthropic and Google can both use uploaded images to improve their models under default consumer settings, and both offer an opt-out, but the opt-out works going forward: it does not pull data out of training runs already started or models already trained, and at Google it does not delete chats already sent to human reviewers. Anthropic’s opt-out is the “Help Improve our AI models” toggle; Google’s is the Keep Activity setting.
- Duck.ai and Venice sit between you and the model provider. Duck.ai removes metadata such as your IP address before forwarding a request, and its providers are contractually barred from using Prompts and Outputs to develop or improve their models and must delete them “at most within 30 days, with limited exceptions”. Neither service says it strips EXIF from the image file itself, so do that yourself. Venice’s own privacy page also says that on its “Anonymous” models you should assume the provider is storing your content.
- Ollama with a vision model and Open WebUI gives you a fully local path. The image never leaves your machine, the only retraining is whatever you do on your own weights, and the privacy work is the same as the local text chat you may already run.
What each hosted chat actually does with your uploaded image
This section compares what each vendor’s own privacy page says. Quotes are taken from the page cited in the source list. Where a vendor does not publish an image-specific statement, that absence is noted - it does not mean the image is handled better.
Anthropic Claude (claude.ai)
Anthropic’s consumer privacy settings cover Claude Free, Pro and Max, including when those accounts use Claude Code; commercial products such as Claude for Work and the API are covered by separate terms. The privacy policy treats content you submit, including content you choose to upload, as your “Inputs,” so the training position is the same for images as for text. Anthropic writes that it “may use your Inputs and Outputs to train and improve Anthropic AI models, unless you opt out through your account settings.”
The opt-out exists and is documented. Anthropic’s help center walks through the path: open claude.ai/settings/profile, then go to Privacy (claude.ai/settings/data-privacy-controls), and toggle off “Help Improve our AI models.” On mobile, the path is name -> Settings -> Privacy. The page is explicit about what the toggle does and what it does not. Anthropic says “we will not use any new chats and coding sessions you have with Claude for future model training,” and that previously stored chats are also dropped from future training runs, but data already included in training that has started, or in models already trained, stays there.
Two exceptions do not respond to the toggle. The privacy policy says that even after an opt-out, Anthropic will use Inputs and Outputs for model improvement when conversations are flagged for safety review, or when you have explicitly reported the materials, for example through the thumbs up/down feedback buttons (a rating may store the entire related conversation). The headline retention language is general: Anthropic keeps personal data “for as long as reasonably necessary,” and deleted conversations are removed from history immediately and from back-end storage within 30 days. The retention help page adds specifics: with model training on, chats may be kept in de-identified form for up to 5 years in training pipelines; chats flagged as violating the Usage Policy are kept for up to 2 years (classification scores for up to 7); and feedback submissions are kept for 5 years. Incognito chats are not used to improve Claude even when the training setting is on.
Net position for images: turn the training setting off, or use an Incognito chat, before you upload anything you would not want in a training pipeline, and do not rate a conversation that contains a sensitive image.
Google Gemini (gemini.google.com)
Gemini Apps (the web app at gemini.google.com and the Android and iOS apps) accept images in prompts. The Gemini Apps Privacy Hub says that when you add an image, Gemini also uses Google Lens technology to understand what is in it and to read the text, and that Google “uses this information just like any other prompt.” Work and school Google accounts may be subject to different data handling terms.
When Keep Activity is on, what you share with Gemini, “like files, videos, screens, and photos,” is saved in your Activity, and Google uses that activity to provide, develop, and improve its services, including training generative AI models, with the help of human reviewers. Data reviewed by service providers is disconnected from your account and kept for up to 3 years, and it is not deleted when you delete your activity. The default auto-delete period is 18 months; 3 months, 36 months, or no auto-delete can be chosen instead. With Keep Activity off, future chats are kept with your account for 72 hours and are not used to train Google’s AI models unless you send feedback. Temporary chats are also kept for 72 hours and are not used to train Google’s AI models. Two limits apply either way: Google says it still uses chats to respond and to protect Google, its users, and the public “including with help from human reviewers” even with Keep Activity off, and that Gemini settings do not control processing of chats to create anonymized data to improve Google services.
Two practical consequences for images. First, “uses this information just like any other prompt” means an OCR-readable screenshot of a sensitive email is treated no differently from a baseline chat. Second, when Keep Activity is on, the path from upload to human review is real and the disconnector is not a guarantee that the image is anonymous - it strips account link, not the content itself.
DuckDuckGo Duck.ai (duck.ai)
Duck.ai is the privacy-fronted chat wrapper. DuckDuckGo’s privacy page says “we call model providers on your behalf, so your personal information (for example, IP address) is not exposed to them.” It says “All metadata that contains personal information (for example, your IP address) is removed before sending Prompts to underlying model providers,” and that providers are required to delete data “at most within 30 days, with limited exceptions for safety and legal compliance.” DuckDuckGo’s stated model providers include Anthropic, Azure OpenAI, OpenAI, and together.ai.
The catch is that DuckDuckGo is a wrapper, not a model, so the underlying provider’s training policy still applies for whatever the provider keeps. DuckDuckGo says its agreements bar providers from “using Prompts and Outputs to develop or improve their models” and require deletion within 30 days, and that “Some models have different levels of privacy protections”. The page names two stronger levels, “Zero Data Retention and Zero Provider Visibility,” and says they are described per chat in that chat’s privacy dialogue. It mentions “files you have uploaded” only in the context of shared links and has no image-specific paragraph, so check the privacy dialogue for the selected model before uploading a sensitive photo.
Venice AI (venice.ai)
Venice routes to a range of vision-capable models, and its docs carry a dedicated Vision section. The privacy docs describe four privacy modes - Anonymous, Private, TEE, and E2EE - and add a critical caveat: “TEE and E2EE are currently available on text models only.” That is, the strongest Venice privacy guarantees do not apply to the image you upload; Venice’s privacy page also lists TEE and E2EE as Pro features. The mode of the model you pick matters more than the policy headline. For Anonymous models, Venice’s privacy page says “Your identity is obscured from the provider, but you should assume the provider is storing your content.” For Private models, the default, it says inference runs on Venice-controlled GPUs or zero-data-retention partner infrastructure, with privacy “enforced by contractual commitments.”
Venice’s main privacy policy defines Prompts as the “text, audio, voice recordings, images, documents, and other materials” you submit, so an uploaded image is a Prompt, and treats Prompts and Outputs as content Venice does not store. It says “We may record your actions (e.g., that you created a chat), but we will not have access to or store the content of your Prompts or Outputs,” and that Venice operates “a zero data retention policy with our model providers”. The policy’s section on likeness uploads (a separate image and video generation feature) says short-term storage of those images is carried out by BytePlus Pte. Ltd. (Singapore). The policy’s zero-retention wording sits uneasily with the privacy page’s warning about Anonymous models, it does not have a separate paragraph on training for chat images, and model providers still process the image to answer. Pick a model marked Private before uploading a sensitive photo, and read both pages yourself before relying on them.
Mistral Le Chat and Proton Lumo
Both come up as image-capable alternatives, but their privacy documents are not covered in this comparison. For sensitive image uploads, the honest position is to assume the same retention and training rules as text until the vendor publishes an image-specific paragraph, and to verify by reading the vendor’s policy directly on the day you upload.
The local fallback: Ollama + a vision model + Open WebUI
A local vision model removes the upload itself from the threat model. The image leaves your machine only over the loopback interface to the local inference server, and no remote service ever receives the pixels.
Ollama’s API documentation supports image inputs. The docs show the images parameter for both /api/generate and /api/chat, accepting base64-encoded images for multimodal models. Ollama’s multimodal blog lists the models it treats as first-class vision targets: Meta Llama 4, Google Gemma 3, Qwen 2.5 VL, and Mistral Small 3.1. The blog is explicit that the multimodal engine was rebuilt because supporting vision models on the legacy llama.cpp path “became more and more challenging,” and now supports document scanning, OCR, and translation.
In practice, the recipe is: install Ollama (and keep to locally run models: Ollama’s FAQ says that when cloud-hosted models are used, Ollama processes prompts and responses on its service, while stating it does not store or log that content; cloud features can be switched off for local-only mode), pull a vision-capable model from Ollama’s library (the multimodal blog lists Meta Llama 4, Google Gemma 3, Qwen 2.5 VL, and Mistral Small 3.1, and the API documentation uses the older LLaVA family as its example), install Open WebUI, and chat through the web UI. Open WebUI passes uploaded images to the local model via Ollama’s API. The image never leaves your machine.
Three honest limits on the local path. First, model quality on image reasoning still trails the largest hosted frontier vision models on hard tasks; OCR and casual image chat are well-served, exotic charts and dense documents less so. Second, you inherit Ollama’s security surface: Ollama’s FAQ says it binds 127.0.0.1 port 11434 by default, but if you bind it to a LAN address without authentication, anyone on your network can hit it. Third, the privacy work you do not skip is still yours: the image is on your disk, the EXIF metadata is still in the JPEG, and the screenshot is still in your clipboard. Local inference is not a substitute for redacting the photo before you decide to keep it.
A pre-flight checklist before any image upload
Use this before uploading any photo to any hosted AI chat.
- Strip EXIF before uploading. Photos carry embedded location, device, and timestamp metadata that travels with the file unless you explicitly remove it. Use your operating system’s photo viewer or a tool like
exiftoolto strip GPS and device fields before the upload. EXIF on its own can leak GPS coordinates down to a few meters. - Decide whether the image has readable text. If it is a screenshot of a sensitive email, an internal dashboard, a tax form, a medical record, or a code editor, the text inside is just text. OCR on the receiving side treats it like a chat. Crop to the relevant area, or upload a screenshot with the URL bar and sidebars blurred.
- Decide whether the image has faces. Faces are biometric identifiers. Even a child in a school uniform is a face. If a hosted chat retains the image for training (whether or not it is disconnected from your account), the face goes with it.
- Pick the chat based on its image position, not its text position. The opt-out toggle that protects your text does not retroactively delete the image you uploaded before you flipped it. Anthropic’s help page says data already included in training runs that have started, or in models already trained, stays there.
- Verify the toggle on the day of the upload. Privacy pages change. Re-read the page that controls your data on the day you upload, not on the day you signed up.
The bottom line
On default consumer settings, Claude and Gemini can use your uploaded images to improve their models. Opt-outs exist at Anthropic and Google, but they are account toggles, not per-upload decisions. Duck.ai is the hosted option that strips request metadata such as your IP address and contracts the downstream provider not to train, with a 30-day delete window. Venice routes to a range of vision models, but its strongest privacy modes (TEE and E2EE) do not yet apply to image inputs, and on its Anonymous models Venice itself says to assume the provider stores your content.
For sensitive image uploads, the honest answer is local: Ollama with a vision-capable model and Open WebUI keeps every pixel on your disk. The local path does not prevent you from uploading your own poorly-reconsidered photo to Google Photos - it prevents the third party from ever receiving it.