Cute AI cats may be fun, but crime scene body bags cross the line

Cute cats dancing a tango on social media may be a bit of fun, but posting AI-generated body bags in a real-life crime scene image defies any common standards of human decency.

The Herald on Sunday’s lead story this week revealed that a Facebook page “dedicated to sharing factual stories sourced from police and trusted news platforms” shared a fake image of body bags being loaded into an ambulance at the scene of an alleged triple homicide in Hastings on April 19.

The incident involved the discovery of a mother and her two young children found dead in their Hawkes Bay home. A 36-year-old man has since been charged with three counts of murder.

The image was posted on a Facebook page called Australia/NZ Crime TV. The post has now been removed. It purported to show a cordoned-off scene with two police cars and two ambulances, into with body bags were being loaded. Quite rightly, the Herald on Sunday chose not to publish the image.

When contacted by the Herald on Sunday a person identified only by their forename said the use of AI was being reviewed and that “our previous use of AI has been limited to generating general graphics that provide visual context to our stories”

The site’s opening title includes the clause: “Some images are altered for legal reasons as investigations are ongoing.” In fact, the use of AI-generated images on the site is extensive and includes the re-rendering of crime scenes. And if that disclosure is expected to warn users of the extensive use of digital fabrication, it falls way short.

In one case, the scene of a fatal stabbing in New South Wales was depicted as an illustration with police interviewing a bystander and photographing a trail of bloody footprints extending along a footpath. A Facebook follower queried the image, which carried no direct form of disclosure.

The response was: “Please note that certain images are made to protect the integrity of crime scenes and specific locations. This post is intended for reporting purposes only, with full respect for the ongoing investigation and those involved.”

Really? The image depicted a viewpoint outside a police cordon, so how could “the integrity of crime scenes” have been an issue? The dominance of bloody footprints looked overly dramatic…and almost certainly was.

Another image (a night-time image apparently from a remote area where human remains were found) pictured a police vehicle whose size is disproportionate to the police officers apparently standing in front of it. The remains were believed to be of a person missing since 2020 but the police forensics tent appeared to  have been erected in the middle of a road. Again, there was no disclosure of image creation or manipulation.

Facebook identifies the page as a “news and media website”. Although none of its content appears to be original reporting, the clear aim is to depict it as journalism. Yet its use of AI imagery is mere fabrication for dramatic effect.

Australia/NZ Crime TV is a stark example of the dangers of using generative AI in news environments. There is no justification for the degree to which it is used on that Facebook page.

Image manipulation is not new. It can be traced from the Cottingley Fairies created by mischievous cousins in 1917, through the ‘removal’ of Stalin’s victims from photographs, to Hasidic newspapers’ removal of Hilary Clinton from the situation room monitoring the killing of Osama bin Laden because their faith did not allow them to picture women in their pages.

Personally, I can recall two examples. On one occasion, a photographer had captured the moment a guide-by-wire anti-tank missile left its launcher. Unfortunately, the guiding wires were obscured, so the photographer drew them in on the print. It wasn’t falsifying the image – the wires had been there – but I did wonder at the time about its efficacy. We might describe it as a line call.

The other was more clear-cut. A photograph from a media conference had a can of soft drink removed because the photographer thought it spoiled the picture. Yet the can was there, bold as brass, in the tv coverage of the media conference and some people asked what had happened to the drink can. Good question: This was no line call, the photographer had altered reality. It was, perhaps, a minor element but the principle was major one: Reality should not be tampered with.

Neither of those examples were serious misrepresentations. Nor were they rarities. A decade ago, 20 per cent of the final entries in the World Press Photo awards were disqaulified because the images were manipulated. None had attempted to deceive the public but their photographs had been ‘enhanced’.

Now, artificial intelligence has the ability to create images that are indistinguishable from reality. The Australia/NZ Crime TV images are amateurish by comparison. There is a clear danger that we will no longer be able to believe what we see.

The Herald on Sunday made the point that there are no laws in this country that specifically require disclosure of AI manipulation or creation. There is, however, a clear ethical obligation and it is not limited to images.

Last week the World Press Photo Foundation tightened its rules to exclude any form of AI manipulation and it defined a “photograph”. That definition now sits at the centre of its entry eligibility and states: “A photograph captures light on a sensor or film. It is a record of a physical moment.” There are further technical requirements that preclude, for example, the very clever real-time manipulation that today’s smartphones can perform. The aim is to reserve the term for a record of that quintessentially momentary slice of time.

Since the invention of photography, our expectation has been that it does, in fact, capture that “moment in time”. It is an expectation that has not always been honoured, but eventually the truth has been revealed.

My camera-wielding colleagues in the past described themselves as “painters in light”. It is an apt description because it was the reflected light that etched their pictures onto a surface. Their skill was in choosing the scene and choosing the moment. They cropped images and made minor adjustments to light and shade but rightly had pride in the fact that the picture showed an audience what had happened in a 25^th of a second or less.

Now, we wonder what is real and what is not. Artificial intelligence is worse than the multiple ‘truths’ of postmodernism on steroids. It can discard reality to create its own universe, one that has advantage of the real world’s large language models without the constraints to which that reality was bound.

For news media, those constraints were professional ethics and a commitment to pursue truth objectively. Those precepts were the basis of the trust that the public resided in journalists. Along with other forms of institutional trust it has been eroded, but trust in the work of journalists must be regained if the public is to have a meaningful grip on reality. And journalism is a craft that most assuredly includes the visual representation of news.

The 2026 JMAD Trust in Media survey from AUT showed that, for the second year running, 60 per cent of the public are uncomfortable with the use of AI in the production of news, even when there is some human oversight. That, and examples like the Australia/NZ Crime TV depiction of a tragic New Zealand crime scene, send a clear message to the real news media.

That message is that artificial intelligence must not be used to alter reality and, where it is employed as a tool to assist newsgathering and presentation, its use must be clearly acknowledged in every instance in which it is employed.

An undertaking buried in hard-to-find codes of ethics will not reassure audiences. And a vague, blanket disclosure that carefully avoids the use of the phrase ‘artificial intelligence’ is no more use than yet another cat video mimicking Fred Astaire and Ginger Rogers.

The image at the top of this column throws reality out the window, and illustrates what can be achieved by ChatGPT.