I recently wrote a short story about a project supported by major tech and media companies aimed at identifying AI-generated or altered content. With the rise of AI-generated text, images, and videos, there has been a growing demand for transparency from lawmakers and internet users. However, adding a simple label to AI-generated content is not as easy as it seems, and current solutions like AI-powered detection and watermarking have significant limitations. As my colleague Melissa Heikkilä has pointed out, most existing technical solutions are inadequate against the latest AI language models. Nevertheless, there is a race to label and detect AI-generated content, which is where the C2PA protocol comes in. Launched in 2021 by the Coalition for Content Provenance and Authenticity, the C2PA provides new technical standards and freely available code to securely label content and provide information on its origins. This includes marking the device used to capture the content, any editing tools applied, and the social media platform where it is uploaded. This information is logged over time, creating a history of the content. The C2PA protocol offers enhanced security compared to other AI-labeling alternatives and can be thought of as a nutrition label for content, ensuring transparency and traceability. The C2PA initially had the support of companies like Adobe and Microsoft, and its membership has grown by 56% in the past six months, including recent adoption by media platform Shutterstock. The protocol operates on an opt-in basis, allowing groups like newspapers or advertisers to add credentials to their media to verify and disclose its origins. The increased interest in and urgency around C2PA can be attributed to the widespread use of generative AI and the expectation of new transparency legislation in the US and EU. However, for the project to succeed, widespread adoption is crucial, and the usability of the protocol is key. The goal is for all major content companies to adopt the standard, ensuring a more transparent ecosystem online. This development is particularly relevant as we approach the US election season, with concerns about AI-generated misinformation.

The race is on to expand functionality and secure partnerships with more social media platforms before the expected influx of AI-generated content. Currently, C2PA primarily focuses on images and videos, but efforts are underway to handle text-based content as well. It's important to note that even with the disclosure of AI use, the harm of machine-generated misinformation may not be fully mitigated. Social media platforms will still need to decide whether to retain such content, and users must exercise their judgment when engaging with and sharing it. This situation is reminiscent of previous initiatives by tech platforms to label misinformation, which have shown mixed results. Ultimately, understanding the impact of social media platforms and their algorithms on political beliefs and civic discourse is an ongoing challenge for researchers. Recent studies on the 2020 election's effects on Facebook and Instagram users' politics reveal a complex relationship, with the segregation of news by political views but limited changes in political beliefs when reshared content is removed from feeds. While these studies have gained attention in the academic world, some scrutiny arises due to their close collaboration with Meta. In other news, there is growing concern about advertising on low-quality content farm sites, and Syracuse's economy could be transformed by a massive investment in computer chips, offering insights into rebuilding the nation's industrial base. The AI Act vote passed overwhelmingly, but the final version may undergo some changes. A Massachusetts bill aimed at limiting police use of technology could establish a regulatory standard for the rest of America, and its possible failure would be a setback for the movement. We apologize for the inconvenience, but we are having trouble saving your preferences. Please refresh the page and update your preferences once more. If the issue persists, please reach out to us at customer-service@technologyreview. com with a list of newsletters you wish to receive.