Forget tweaking thigh lines, removing wrinkles or transposing faces: by adding generative AI to Photoshop, Adobe has guaranteed that you will never be able to trust the authenticity of any image you see, ever again.

Built into the latest beta version of Photoshop, Adobe’s Generative Fill tool allows photographers to not only adjust photos they have taken – a much-maligned practice that long ago turned the software’s name into a verb – but to tweak and add entirely new elements to their photos using English language prompts.

The feature leverages Firefly, a generative AI engine that Adobe has trained on a large range of stock photos, openly licensed images and public domain content that is no longer covered by copyright.

Firefly also powers other Photoshop features including generating complete images, stylising text by using descriptive prompts, recolouring images, turning sketches into images, creating images from 3D models, recomposing photos, and more.

Yet Generative Fill may be the final nail in the coffin for the integrity of photography: by not just creating new images – but allowing existing images of people and things to be seamlessly blended into new environments – the tool enables artists and propagandists alike to create convincing images involving real people, places, and events.

Galleries show artists using the tool with famous album art, paintings and memes to put the images in entirely new contexts – such as a young boy photographing the Mona Lisa on his smartphone, or Katy Perry sleeping on a large, fluffy, pink cat.

Yet by allowing photographers and artists to play freely and loosely with the truth, the technology also risks shattering assumptions about the integrity of images – as when one tinkerer used Generative Fill to put the famous Pope in a puffer jacket (itself an AI fabrication) into a streetscape while carrying luggage and wearing leather pants.

When deepfakes become real

The integration of AI into photo editing software is nothing new – AI techniques have long-powered selection, object removal and other tools, with Skylum launching an AI-based spinoff of its Luminar image editor in 2020 – but the addition of generative AI capabilities means the technology can now not only enhance photographic truths, but create new ones.

“Now that we are entering a new era of AI, the advent of generative models presents a new opportunity to take our imaging capabilities to another level,” Adobe vice president for Photoshop and Lightroom product management and strategy Pam Clark said in announcing the new features in Photoshop’s May update.

Calling the new feature “a new chapter in our history”, Clark said the new tools would empower content creators “to work at the speed of their imagination”.

Seamless integration of generative AI into the widely-used photo editor is also all but certain to spawn a flood of potentially injurious deepfakes – which have already raised eyebrows as privacy advocates, cyber security experts, journalists, and others lament the death of verifiable written, visual and even voice communications.

It may not threaten human life in the way that former Google head Eric Schmidt recently warned that AI would, but with Google recently partnering with Adobe to integrate Firefly into its Bard generative AI platform, Generative Fill will hasten the death of verifiability.

While the industry tries to wrest back control of image verification through efforts like the Adobe-backed Content Authenticity Initiative (CAI) – an element of the Coalition for Content Provenance and Authenticity (CCPA) that will stamp images with details of which content has been modified using Generative Fill – broader access to generative AI tools risks overwhelming efforts to control its use.

Fluid realities may be here to stay, however, with Accenture’s recent Technology Trends 2023 report saying that generative AI will drive “the next wave of business transformation” and the creation of “a shared reality that seamlessly combines our physical lives of atoms and our digital ones of bits.”

“Many people are using [generative AI] to generate purely digital images and content,” the firm notes, “but we already see how it will shape the future of science, enterprise data, how we design and manufacture products, and so much more.”