How invisibility cloaks could make us disappear – at least from AI

The desire to disappear has been strong throughout history. It didn’t go well for the protagonist in H. G. Wells’s The Invisible Man, but that is because his invisibility was permanent. What was needed – and what was longed for – was a means of disappearing temporarily, as popularised by Harry Potter’s invisibility cloak.

Metamaterials developed in the early 21st century gave hope that a garment offering universal invisibility was feasible. But while some forms of cloaking device did become possible, the sheer level of engineering required to produce them meant they remained rare, ultra-expensive and out of reach to the vast majority. (Nor was the fabric anything like that of the gossamer cloaks of wizardry.) Instead, invisibility went down a different path. Clothes were designed that concealed their wearers not from other people, but from a more insidious enemy: artificial intelligence. They didn’t visibly disappear, but their identity – even their humanity – was concealed from ubiquitous visual-recognition systems.

Metamaterials are engineered fabrics containing nanostructures or microstructures that control and manipulate electromagnetic waves’ paths. Like water flowing around a stone, when light hits a metamaterial, it isn’t absorbed or reflected but redirected.

The main issue with metamaterials was that they were specific to wavelength. Some of the first invisibility devices, such as those conceived by John Pendry at Imperial College London in 2006, could conceal objects, but only from microwave radiation. Hiding from shorter wavelengths, including visible light, required an even more highly engineered nanostructure material.

One promising approach came from an engineered optical device called a metalens, which is similar to a traditional lens in that it can manipulate light, but is flatter and thinner. By pairing metamaterials with metalenses, scientists could construct fabrics that manipulated light such as to render an object or person behind it invisible. Still, fabrication was too hard for it to go mainstream.

Key here were materials first made in 2024 using self-adaptive photochromism (SAP) (SAP) – essentially the same method used by octopuses and chameleons to change the colour of their skin to match their background. They contained molecules that change structure when exposed to light, taking on the colour of the background. “Chameleon clothing” became common among field biologists, who were able to observe animals without detection, and of course among the military. But fabrics with changing colours became immensely popular among fashion designers too.

It was in the early 2030s when SAP clothing was combined with electronics that could dynamically manipulate and program patterns, that a new kind of invisibility was discovered. It wasn’t long before mainstream fashion acquired a more radical political aspect than had been seen for many years.

In 2024, students at Wuhan University in China had developed InvisDefense, a fabric that rendered wearers invisible to cameras run by AI. The key lay in the pattern, designed to disrupt and evade image-recognition systems. When caught on CCTV, a person wearing InvisDefense clothing wasn’t classified as human by the AI.

But the patterns in InvisDefense clothing were static. Dynamic SAP materials then came along that could be programmed to display a morphing, transient, endless swirl of colours. AI systems couldn’t recognise so-called polymorph clothing or even classify the wearers as people – the systems simply categorised them as noise.

It was perhaps not surprising that InvisDefense was developed in China. By the 2020s, the citizens of China were among the most surveilled in the world. (China had some 200 million camera systems in the 2020s, while the UK had around 7.5 million cameras and the US some 50 million.) It was hard to argue that CCTV didn’t play a protective role, but it also played a stifling, authoritarian one.

The next SAP development was ultra-thin full-face polymorph masks, known as polymasks, designed to be worn with regular clothing. The masks created a look that appeared to be completely authentic, moving naturally with the muscles beneath. However, the outward appearance was completely unlike the real face beneath.

Governments initially tried to regulate access to polymasks, but the technology required to produce the material was relatively simple and restricted access became impossible. There was, inevitably, a criminal element among the users of polymasks, but most people used the mask to escape relentless targeted advertising, racial profiling and the endless surveillance of the modern world.

Rowan Hooper is New Scientist‘s podcast editor and the author of How to Spend a Trillion Dollars: The 10 global problems we can actually fix. Follow him on Bluesky @rowhoop.bsky.social