How to block AI image generators

AI image generation has made a specific class of harmful content easier to access than ever before. This post covers why it is worth blocking, and provides a practical blocklist for Cold Turkey, DNS tools, and iPhone profiles.

Why AI image generators are a unique problem

Traditional content blocking has always been about blocking existing content — websites, images, and videos that already exist somewhere and can be found. AI image generators change this: they create content on demand, from a text prompt, with no moderation delay and no existing record that can be filtered.

Many of the most widely used generators offer little to no content filtering for sexual, violent, or otherwise harmful imagery. Some are designed from the ground up to produce it. Even tools that claim to filter content often have well-documented bypass methods that require minimal effort to use.

The result is that anyone with access to these websites can generate harmful imagery within seconds, with no trace and no external record. For people trying to limit their own access, or for parents and accountability partners trying to set up effective restrictions, this makes AI image generators a significant and growing loophole.

The blocking challenge

Blocking AI image generators is harder than blocking social media or streaming sites for two reasons. First, the category is expanding quickly — new tools launch regularly, and many established platforms (search engines, productivity tools, chatbots) are quietly adding image generation features. Second, many generators are hosted on research and developer platforms like Hugging Face or Replicate, which also host legitimate tools — making broad domain blocks less practical.

The most reliable approach is to maintain a specific blocklist targeting the most-used standalone generators, combined with a broader policy of blocking developer platforms if they are not needed for work.

Blocklist: AI image generation websites

The following list covers the most widely used AI image generation platforms as of mid-2026. It is intended for use with Cold Turkey on Windows, or at the DNS or profile level for iPhone via tools like Techlockdown.

What to do about platforms you cannot easily block

Some image generation is embedded deeply into tools that are otherwise necessary. ChatGPT's image generation (via DALL-E), for example, is part of the same domain as the text interface. Blocking OpenAI entirely removes access to the text tool as well.

There is no clean technical solution for this. The practical options are to accept that access exists and rely on accountability, or to switch to a text-only interface such as the API or a client that does not expose the image feature. If ChatGPT image generation is a specific risk, it is worth noting in your accountability setup so that a trusted person knows to ask about it.

Keeping the blocklist current

New AI image tools are launching continuously. The list above will become incomplete over time. Two habits help keep the blocklist relevant:

• Review it monthly and add any new tools you have become aware of — including ones you were tempted to use.
• Block broad categories where possible rather than individual sites. Blocking entire developer infrastructure platforms (Hugging Face, Replicate) is blunt but effective if those platforms are not needed for work.

A blocklist that is 90% complete and actively maintained is more useful than one that is theoretically comprehensive but never updated.

Back to blog