AI-generated media is challenging newsroom verification. Many authentication tools lack direct journalist input. New research highlights gaps in current solutions and urges more collaboration.
News organizations are facing mounting challenges as AI-generated media makes it easier and cheaper to create convincing fake photos, videos, documents, and even audio. This surge in synthetic content is putting pressure on newsroom verification processes, with many tools and initiatives failing to meet the practical needs of journalists.
In June, a group of journalists, technologists, and academics convened at NYU’s Arthur L. Carter Journalism Institute to address the growing threat to media authentication. The workshop, organized by Princeton’s Center for Information Technology Policy (CITP), led to a new report that synthesizes expert findings on the fight for facts in the AI era. The report notes that while AI can enhance verification, it also fuels an arms race with fabricated information flooding digital platforms.
Despite a wave of new verification products and initiatives, the report finds that many are developed without sufficient input from working journalists. One prominent example is C2PA, or the Coalition for Content Provenance and Authenticity, launched in 2021 by Adobe, Microsoft, the BBC, and others. C2PA embeds cryptographically signed manifests in media files to track their origin and history. However, newsroom adoption has been hampered by technical obstacles. Some photo editing software strips these manifests, and most content management systems (CMS) do not support C2PA, resulting in lost provenance data when images are published.
Even major news organizations like the Associated Press and The New York Times have encountered issues with C2PA implementation. The report cites a newsroom where the manifest was removed during editing, and the publisher’s CMS failed to retain it for public-facing images. This raises questions about whether audiences can trust a publisher’s chain of custody if transparency is incomplete. Social media platforms further complicate matters by stripping metadata, including C2PA information, from shared images.
Safety concerns also persist. Cryptographic signatures may expose journalists and sources by attaching identifying details to media, posing risks in authoritarian environments. Human rights group Witness has highlighted these dangers in recent research.
Other authentication tools, such as AI image and video detectors, present their own challenges. Many provide confidence ratings-such as an 80% likelihood that an image is AI-generated-but these scores are often unhelpful for reporters and confusing for audiences, who typically want a clear answer on authenticity. The report points out that these tools often function as black boxes, making it difficult for journalists to explain or trust their results.
Given these obstacles, the report urges technologists and journalists to collaborate more closely in developing and testing authentication tools. The authors recommend that reporters be involved throughout the process to ensure solutions are practical and effective for newsroom workflows.
These findings echo broader concerns about the impact of AI on media trust and verification. For example, recent coverage of publisher strategies to address AI-driven content scraping, such as UK publishers moving to charge AI firms for unauthorized content use, highlights the ongoing tension between technology and editorial control.
The full CITP report offers additional recommendations and a detailed overview of how AI is reshaping verification work across the news industry.
