Meta, the company that owns Facebook, Instagram, and Threads, has announced plans to label all images on its platforms that have been created using artificial intelligence (AI). This move comes as Meta acknowledges the increasing importance of AI-generated images in the digital space, particularly during a year with significant elections worldwide.

Previously, Meta had been labeling photorealistic images created using its own Meta AI feature with "Imagined with AI" labels. However, the company is now developing "industry-leading tools" that will enable it to identify invisible markers on AI-generated images sourced from other platforms like Google, OpenAI, Microsoft, or Adobe. Meta intends to roll out these labels on Facebook, Instagram, and Threads in the coming months.

Sir Nick Clegg, Meta's president of global affairs and former British deputy prime minister, emphasized the significance of this move in a statement. He noted that the distinction between human and synthetic content is becoming increasingly blurred, and the company aims to learn more about how people create and share AI content. The insights gained will inform industry best practices and Meta's approach moving forward.

Meta has been collaborating with industry partners to establish common technical standards for identifying AI content. The company's technology will detect "industry standard indicators" to label AI-generated images, ensuring that the labels are available in all languages.

The decision to prioritize AI-generated images and videos stems from the exponential growth of such content on the internet. Since 2022, an estimated 15 billion AI-generated images have been uploaded online. While many of these images are harmless and fall into categories like cute kittens, sci-fi, or anime, a significant number are harmful. Examples include fake explicit images of individuals, uploaded without their consent, and politically motivated misinformation aimed at manipulating the truth.

The urgency to address this issue also arises from legislative measures and growing pressure on social media companies. The UK implemented the Online Safety Act in 2022, criminalizing the uploading of fake explicit images without consent. In the US, lawmakers recently criticized social media platforms for failing to ensure user safety and called for laws to compel them to do more.

Meta's announcement is expected to have a ripple effect, compelling its competitors to adopt similar labeling practices. However, the effectiveness of these labels remains uncertain. Research teams have demonstrated that digital watermarking, even when buried in image metadata, can be easily removed. Meta acknowledges that its technology is not flawless. The true test will be whether the proliferation of harmful fake images decreases in the coming months, which may prove challenging given the vast number of users across Meta's platforms.

If the expected decline in harmful content does not materialize, the next challenge may involve legal action against major tech companies. Some experts argue that hitting big tech in their pockets through lawsuits will be the catalyst for real change in this area.

Sir Nick acknowledges that Meta cannot currently identify all AI-generated content, as creators can strip out invisible markers. However, the company is actively working on developing classifiers to automatically detect AI-generated content, even without these markers. Simultaneously, Meta aims to make it more difficult to remove or alter invisible watermarks. Sir Nick emphasizes the adversarial nature of the AI space and the need for constant innovation to stay ahead of those who seek to deceive with AI-generated content.

In addition to labeling AI-generated content, Meta plans to introduce a feature that allows users to disclose when they are sharing such content. This disclosure will enable Meta to add appropriate labels to the content, providing transparency to users.

The controversy surrounding AI-generated images has been evident in recent months. Deepfake images of pop superstar Taylor Swift, believed to have been created using AI, circulated widely on social media in January. These images were so realistic that users struggled to distinguish them from real photographs. The dissemination of sexually explicit images of Swift prompted concerns, with the White House calling on social media companies to enforce their own rules and urging Congress to legislate on the issue.

Similarly, in the UK, a slideshow of eight images purportedly showing Prince William and Prince Harry at the King's coronation gained significant traction on Facebook in 2023. One of the images depicted an emotional embrace between the brothers, seemingly reconciling after reports of a rift. However, these images turned out to be fake, highlighting the potential for AI-generated content to deceive and manipulate public perception.

Meta's decision to label AI-generated images represents a significant step toward addressing the challenges posed by this technology. However, the effectiveness of these labels and the broader issue of harmful fake images online remain complex and multifaceted problems that require ongoing vigilance and innovation.