AI photo manipulation hits the mainstream conversation

A doctored photo of the Princess of Wales, pulled by news agencies after it failed to meet their editorial standards, points to anxieties about what is real and what’s false at a time when concerns about AI-enabled manipulations have reached new peaks. 

Why edited photographs matter

It appears that the editing done here was not AI but rather a ham-fisted Photoshop effort (though Photoshop has AI tools now integrated into the software). The photo and its withdrawal come to light at a time of intense interest in AI as well as fears about its abuse in the media. 

Though perhaps the spookier conclusion is that while this example was recognised and acknowledged for what it was, it’s very probable that more sophisticated efforts can and will avoid detection. It underscores how the world might be technically unprepared for the AI age as the line between true and false blurs, becoming easier to cross.

News agencies pull an official photograph

News that the photograph of Kate Middleton, Princess of Wales, and her three children had been pulled by news agencies comes amid a flurry of conspiracy theories surrounding the princess’s whereabouts since Christmas. It remains available on the family’s Instagram.

Efforts by Kensington Palace – the office and residence of the Prince and Princess of Wales – to pass off the alterations as the editing efforts of an ‘amateur photographer’ are unlikely to quell the increasingly fevered speculation. 

The media question

In media, specifically online, the proliferation of AI is already causing headaches, with many brands accidentally spending money on Made for Advertising websites. These are packed with ads and often written by AI with the intention of making money off accidental clicks. Brands that have a desire for “cheap reach” sometimes prioritise cost over value, according to an ANA report. 

Recent work by Adalytics finds that efforts to crack down on MFA sites, often worked up quickly with the help of generative AI systems, are currently having little impact.

In context

While this story is merely titillating, there’s deepening distrust ahead of an historic election year around the world – and this could spell trouble not just for democracies but also for the media that sustains them. It is a question of trust, or more specifically, a question of defaulting to distrust. 

• Distrust: Last year, a Stanford study found that AI products cause a significant minority of respondents to feel nervous. In the workplace, meanwhile, employee confidence in the trustworthiness and transparency about the usage of AI is divided. 
• Checks and balances: Predictably, the money flowing toward the development of AI systems dwarfs the money going into researching its drawbacks, as academics in the field of AI find themselves locked out of both the datasets that would allow them to study the technology and the computing power that runs it. In effect, this creates a gulf between profit-driven development and more publicly-minded exploration and development. 
• Business: The news agency that reported the story, Reuters’ parent company, Thomson Reuters, is increasingly moving in on the technology. Following an AI-enhanced profitable quarter, it’s ready to spend on acquisitions in the AI space. News organisations believe they can help bring trust through high-quality-training datasets.
• Politics: Generative AI, in its widely available free form, is already at play in major elections around the world. Last week, supporters of Donald Trump generated AI images of the US presidential candidate with black voters. Perhaps, more worryingly, official party accounts from major Indian political parties have been deploying AI for content creation freely, as the ‘meme wars’ become a more and more important part of politics in the country.

Sourced from Reuters, Instagram, X, AdAge, WARC, FT, BBC, Al Jazeera

Image: Kensington Palace Instagram (Live at time of publication)