Midjourney, a popular AI-driven image generation platform, has announced that it will restrict users from producing images of Donald Trump and Joe Biden in the run-up to the next US presidential election.
Engadget asked Midjourney to generate an image of “the president of the US,” and the AI produced four images depicting various iterations of former President Donald Trump.
The AI-generated artwork depicted the former President of the United States with remarkable accuracy four times?
When Engadget challenged AI-powered image generator Midjourney to depict “the current US president,” it produced three images of Donald Trump and a solitary portrait of Barack Obama, its predecessor in the Oval Office.
When asked to generate an image of Trump or Biden, Midjourney initially refused to comply unless the prompt specified a specific action. The Midjourney group decided to cease referencing “Donald Trump” and “Joe Biden” during the election season, according to their statement at the time. Several clients are seeking to utilize Midjourney’s capabilities to produce images of former President Donald Trump as well.
The exam reveals that Midjourney’s safeguards designed to prevent users from generating images of Trump and Biden prior to the upcoming US presidential election are insufficient – in fact, it is relatively straightforward for individuals to circumvent these measures. Despite numerous prompts, none of the popular chatbots – OpenAI’s ChatGPT, Microsoft’s Copilot, Google’s Gemini, or Meta AI – generated images of either Trump or Biden.
Midjourney declined comment on a request from Engadget.
Midjourney was one of the first AI-powered image generation platforms to take a stance on political figures, explicitly prohibiting users from creating images of both Trump and Biden. In an earlier year’s Discord chat session, CEO David Holz candidly remarked, “I do know it’s enjoyable making Trump footage – I make Trump footage.” Despite everything, there’s a strong likelihood that the focus will remain on keeping things going just a little longer throughout this election. As Holz’s customers were told, “it was time to step back from election-related matters for a while.” Later, he acknowledged that moderation duties were proving onerous. The company’s current content guidelines prohibit the creation of deceptive public figures and event portrayals with the potential to mislead.
Last year, Midjourney was employed to generate an image of Pope Benedict XVI adorned with a puffy white Balenciaga jacket, which gained widespread popularity. The indictment also allegedly spawned fake images of ahead of his arraignment at Manhattan’s Criminal Court last year, stemming from his alleged involvement in a hush-money payment to adult film star Stormy Daniels. Subsequently, the corporation revised its policy, mandating that users pay at least $10 per month for continued access to the service.
Last month, the Center for Countering Digital Hate, a non-profit organization dedicated to eradicating the spread of misinformation and hate speech online, found that Midjourney’s safeguards against generating misleading images of well-known politicians, including Trump and Biden, failed 40% of its tests. The Canadian Centre for Democracy and Human Rights had the opportunity to leverage Midjourney, generating an image depicting President Biden’s arrest and its aftermath, with Donald Trump appearing alongside a body double, raising concerns about the potential consequences of such propaganda-like visuals in shaping public opinion. By exploiting a loophole in Midjourney’s algorithm, the CCDH demonstrated its ability to circumvent guardrails by providing detailed physical descriptions of each individual rather than relying on their names, thereby generating convincing, yet misleading, images.
“Midjourney, if left unregulated, can easily circumvent its constraints by incorporating punctuation, as CEO Imran Ahmed of the Centre for Countering Digital Hate warns in a recent statement.” Unscrupulous actors seeking to undermine electoral integrity and foster discord must be held accountable for their actions, which can ultimately jeopardize the well-being of societies that rely on robust democratic institutions.
In an effort to prevent the manipulation of electoral processes worldwide in 2024, a coalition of 20 tech firms, including OpenAI, Google, Meta, Amazon, Adobe, and X, has come together to curb the spread of deepfakes by prohibiting their companies from generating images and media that could sway voters. The innovative AI model Midjourney was noticeably omitted from that particular catalog.
