In a significant shift in policy, Meta, the tech giant behind Facebook and Instagram, announced on Friday that it will begin labeling AI-generated media starting in May 2024. The decision comes as a direct response to growing concerns over the potential misuse of deepfakes and manipulated content, particularly in the context of elections and public discourse.
Under the new policy, Meta will no longer remove manipulated images and audio that don’t violate its Community Standards. Instead, the company will rely on a combination of labeling and contextualization to provide transparency and additional context to users. The “Made with AI” labels will cover a wide range of AI-generated content, including video, audio, and images. Furthermore, content deemed at high risk of misleading the public will receive a more prominent label to alert users of its potential to deceive.
The changes follow criticism from Meta’s Oversight Board, an independent body tasked with reviewing the company’s content moderation decisions. In February, the board urged Meta to urgently overhaul its approach to manipulating media, citing the rapid advancements in AI technology and the ease with which highly convincing deepfakes can be created. The board’s warning came amid growing fears of rampant misuse of AI-powered applications for disinformation on social media platforms, particularly in a pivotal election year not only in the United States but worldwide.
Meta’s decision aligns with an agreement made in February among major tech giants and AI players, including Google and OpenAI, to combat manipulated content intended to deceive voters. As part of this agreement, the companies committed to using a common watermarking standard to tag images generated by their AI applications, making it easier to identify and track the spread of manipulated content across platforms.
The rollout of Meta’s new labeling system will occur in two distinct phases. AI-generated content labeling will begin in May 2024, while the removal of manipulated media solely based on the old policy will cease in July. Under the new standard, manipulated content will remain on the platform unless it violates other Community Standards, such as those prohibiting hate speech or voter interference. This approach aims to strike a balance between protecting free speech and ensuring that users are aware of the nature of the content they consume.
The announcement comes amid growing concerns over the potential misuse of AI deepfakes, with recent examples highlighting the ease with which such manipulations can be created and disseminated. In one instance, a manipulated video of US President Joe Biden inappropriately touching his granddaughter circulated online, while in another, a robocall impersonation of Biden urged voters not to cast ballots in the New Hampshire primary. The issue is not limited to the United States, as evidenced by the case of former Pakistani prime minister Imran Khan’s party using AI to generate speeches from their jailed leader.
As AI technology continues to advance at an unprecedented pace, the debate over the balance between free speech and the need to combat disinformation and manipulated content is likely to intensify. Meta’s new labeling system represents a significant step in addressing these concerns, but the effectiveness of the approach remains to be seen. Critics argue that labeling alone may not be sufficient to curb the spread of deepfakes and manipulated content, while others worry that the policy change could have a chilling effect on free speech.
Moving forward, it will be crucial for Meta and other tech companies to engage in ongoing dialogue with policymakers, experts, and the public to refine their approaches to AI-generated content and ensure that the benefits of this transformative technology are harnessed while minimizing its potential for harm. As the world enters a new era of AI-driven communication, finding the right balance between innovation, transparency, and responsibility will be key to fostering a healthy and informed public discourse.