DALL·E 3 is now available in ChatGPT Plus and Enterprise

OpenAI has flipped the switch on DALL·E 3 for anyone paying for ChatGPT Plus or using the Enterprise tier. Instead of jumping to a separate tool, you can now ask ChatGPT to generate an image mid-conversation and it just does it, refining the picture the same way it refines a paragraph of text if you push back and ask for changes.

The bigger story here isn't the feature drop itself but what OpenAI did before flipping it on. The company built out what it calls a safety mitigation stack specifically to get this model ready for a much larger audience than the research previews DALL·E 3 has seen so far. That's a meaningful shift in posture: image generators have a nasty habit of getting abused for deepfakes, harassment, and misinformation, and OpenAI clearly wanted guardrails in place before millions of Plus subscribers started typing prompts.

OpenAI is also talking, in general terms, about provenance research — the work of figuring out where a given image actually came from and whether it was AI-made. This matters more every month. As synthetic images get harder to distinguish from real ones, the ability to trace origin becomes less of a nice-to-have and more of a basic trust requirement for the internet at large. OpenAI isn't laying out a finished system here, just signaling that the research is ongoing and that they see it as part of the responsible-release package.

What's notably absent is detail. There's no breakdown of exactly what the safety stack filters for, no numbers on rejection rates, no specifics on how the provenance tracking would actually work in practice. OpenAI is asking users to trust that the guardrails exist without showing much of the plumbing, which is a pattern this company repeats a lot: ship the product, gesture at the safety work, save the granular details for a future paper or blog post.