What Journalists Should Know About Deepfake Detection in 2025
By Columbia Journalism Review / Tow Center.
Summary: Non-technical guide for journalists on detecting synthetic media and deepfakes in reporting workflows.
Relevance: Falls under Surveillance & Civic Life: a practical guide to the limits of deepfake-detection tools that applies just as much to social workers who encounter AI-manipulated audio, video, or images as evidence in custody disputes, intimate partner violence cases, or client disclosures.
Key findings: Detection tools like Deepware Scanner and Hive Moderation return ambiguous outputs (e.g., “70 percent human”) without explaining what was altered, and accuracy “drops sharply” outside the datasets they were trained on, especially for audio. A University of Mississippi study found journalists over-trusting detection results that confirmed their existing biases; the guide’s core recommendation is to treat any single tool’s output with “a high degree of skepticism” and never substitute it for human verification.
Read the original: https://www.cjr.org/tow_center/what-journalists-should-know-about-deepfake-detection-technology-in-2025-a-non-technical-guide.php
