The problem is simple: it’s hard to know whether a photo’s real or not anymore. Photo manipulation tools are so good, so common and easy to use, that a picture’s truthfulness is no longer guaranteed.

The situation got trickier with the uptake of generative artificial intelligence. Anyone with an internet connection can cook up just about any image, plausible or fantasy, with photorealistic quality, and present it as real. This affects our ability to discern truth in a world increasingly influenced by images.

I teach and research the ethics of artificial intelligence (AI), including how we use and understand digital images.

Many people ask how we can tell if an image has been changed, but that’s fast becoming too difficult. Instead, here I suggest a system where creators and users of images openly state what changes they’ve made. Any similar system will do, but new rules are needed if AI images are to be deployed ethically – at least among those who want to be trusted, especially media.

Doing nothing isn’t an option, because what we believe about media affects how much we trust each other and our institutions. There are several ways forward. Clear labelling of photos is one of them.

Photo manipulation was once the preserve of government propaganda teams, and later, expert users of Photoshop, the popular software for editing, altering or creating digital images.