ChatGPT Images 2.6 gives VAs two new tools for client work: GIFs and saved reference images. The trap is that both make it very easy to put a client's photos and logo into an AI tool before anyone has said yes.
Here's what changed and what I'd sort out with a client first.
What's New in ChatGPT Images 2.6
ChatGPT Images is the picture maker inside ChatGPT. OpenAI's November 4 release notes say the new version is faster and better at keeping details the same across edits. Their example is a character's outfit staying put when you change the background.
That's the upgrade I care about for client graphics. If you've ever asked AI to change one small thing and got a whole new picture back, you know why. OpenAI's help page still warns that a selected edit can spill past the area you highlighted, so check the whole image each time.
OpenAI's system card for ChatGPT Images 2.6, published November 4, 2026
GIFs and Saved Reference Images
A GIF is a short image that moves on a loop. You ask ChatGPT for a looping GIF and describe what should move, or you upload a still image and ask it to animate that. Then you can change the speed, save it or share it.
GIF creation is still rolling out, so you may not see it yet. And ChatGPT can't edit an animated GIF you upload, only the ones it made.
My ideas for VA work: a moving header for a client newsletter, a waving mascot for Instagram Stories, or a small looping sale badge.
Saved reference images save you from uploading the same photo in every chat. You upload it, ask ChatGPT to save it as a reference and give it a name. Later you type @ and that name in your prompt. OpenAI's example is a dog called Pepper. For us, that could be a product photo saved as @BlueMug.
[MAYE (optional): make one client-style graphic and one GIF from a made-up brand, then add them here with a line on what you had to fix.]
Who Gets It and What It Costs
OpenAI's help page says ChatGPT Images is on all tiers, on web, iOS and Android. So yes, the Free plan can make images.
OpenAI's pricing page, in US dollars, lists Free ($0) with limited image creation, Go ($8 a month) with more, Plus ($20) with more complex and accurate images, and Pro (from $100) with unlimited, faster creation, subject to abuse guardrails. It doesn't give an image count for Free, Go or Plus, and neither does the help page.
Plus, Pro and Business also get an extra mode called Images with thinking, with Enterprise and Edu coming soon. The pages I checked don't say whether GIFs or saved references differ by plan.
The Permission Trap
OpenAI's terms say that between you and OpenAI, you own what you make. But you promise you have the rights to what you upload, and you can't use the service to violate anyone's rights. The same terms warn that output may not be unique, so someone else could get something similar.
I'm not a lawyer, so here's how I'd handle it in practice.
Logos: don't ask AI to redraw a client's logo, because a redrawn logo is a wrong logo. Make the graphic without it, then add the real logo file in Canva. Leave other companies' logos and famous characters out completely.
Photos: ask before you upload, and ask in writing. If a staff member or customer is in the photo, the client needs their okay too. OpenAI's usage policies ban using someone's likeness without consent in ways that could confuse people about what's real.
Saved references need one extra question, because the photo stays ready in your account. The help page doesn't explain how to delete one, or how long it's kept. Find that out before you save anything for a client.
On individual plans, OpenAI may use your content to train its models unless you turn off "Improve the model for everyone" under Settings, then Data controls. That only covers new chats, and a thumbs up or down can still send that whole chat to training. Business plans aren't used for training by default.
Last, honesty. OpenAI's terms say you can't present AI output as human-made. Its system card, the safety report for this release, says OpenAI uses C2PA metadata (a hidden label showing where a file came from) and is adding SynthID, an invisible watermark. So don't let a client think a designer made it.
What This Means for VAs
If you manage a client's social media or newsletter, this gives you quick visuals for the small jobs that never reach a designer: a story graphic, a sale badge, a moving header. Brand work and anything legal or medical should still go past a human who owns that decision.
Before you offer it, add three lines to your client notes. Which photos and logos can go into AI tools? Who approves each graphic before it goes out? Does the client want AI-made visuals named as such?
I wrote more about that talk in setting AI expectations with clients.
Then practise on a made-up brand, so no client files are involved while you learn. My 30 free VA prompts will help you write clear requests.
What's the first GIF you'd make for a client?
