Skip to content
Longterm Wiki
Back

MIT Media Lab: Detecting Deepfakes

web

Relevant to AI safety discussions around misuse of generative models; this project represents a public-facing, human-centered approach to mitigating deepfake harms rather than a purely technical solution.

Metadata

Importance: 42/100tool pageeducational

Summary

MIT Media Lab's Detect Fakes project investigates how people can identify AI-generated media, particularly synthetic video and audio. The project uses an experimental website to test and train public ability to spot deepfakes through critical observation techniques. It aims to raise awareness and build human-level media literacy as a defense against AI-generated disinformation.

Key Points

  • Develops interactive tools to help general audiences distinguish real media from AI-generated deepfakes through hands-on experience.
  • Focuses on human perceptual skills and critical observation as a complement to automated detection methods.
  • Addresses the societal risks of synthetic media by building public awareness and media literacy.
  • Part of broader MIT Media Lab research into the trustworthiness and authenticity of digital content.
  • Highlights the growing challenge of content verification as generative AI capabilities advance rapidly.

Review

The Detect Fakes project by MIT Media Lab addresses the growing challenge of AI-generated media manipulation by developing strategies to help ordinary people critically evaluate digital content. By creating an interactive website and providing detailed guidelines, the researchers aim to enhance public understanding of deepfake technologies and their potential risks. The project's methodology involves exposing users to curated deepfake and authentic videos, teaching them to recognize subtle computational manipulations through eight key observation points. These include analyzing facial features, skin texture, eye movements, lighting, and lip synchronization. While the approach doesn't rely on advanced machine learning algorithms, it emphasizes human perception and critical thinking as essential tools in combating misinformation, representing an important complementary approach to technical deepfake detection methods.

Cited by 2 pages

Cached Content Preview

HTTP 200Fetched Apr 9, 20266 KB
Overview ‹ Detect DeepFakes: How to counteract misinformation created by AI — MIT Media Lab 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 

 
 

 

 

 
 

 
 
 
 

 
 
 

 
 
 
 
 
 Find People, Projects, etc. 
 

 

 
 
 
 Search 

 
 
 
 

 
 
 

 

 

 
 
 
 

 
 
 

 
 
 
 

 

 
 
 

 
 
 

 
 
 Login

 Register

 

 
 
 

 

 

 

 
 
 Email: 
 
 

 
 
 
 

 
 
 Password: 
 
 

 
 
 
 

 

 

 
 
 

 
 Work for a Member organization and need a Member Portal account? Register here with your official email address.

 Register a Member Account 
 
 

 
 

 
 

 
 

 

 
 
 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 Project 

 
 Detect DeepFakes: How to counteract misinformation created by AI

 
 

 

 

 
 
 
 
 
 
 

 
 

 
 
 
 
 Creative Commons
 

 
 
 
 Attribution 4.0 International 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 Matt

 

 

 

 
 
 

 
 

 
 
 

 

 
 
 
 
 
 
 
 
 

 
 
 
 Project Contact:

 

 
 
 
 
 Matt Groh
 
 

 
 
 
 
 
 
 
 
 matthew.groh@kellogg.northwestern.edu 
 
 

 
 
 
 
 
 
 
 Project Website 
 
 

 
 
 
 
 
 
 
 
 
 
 

 
 
 

 
 
 
 Other Press Inquiries 
 

 
 
 

 
 
 Groups

 
 
 
 
 
 
 
 
 
 
 

 

 
 
 

 

 
 

 

 
 
 
 
 
 
 

 
 
 

 
 
 
 
 
 
 
 
 
 Detect DeepFakes: How to counteract misinformation created by AI was active from April 2020 to January 2025
 
 

 

 

 
 

 
 See for yourself how accurately you can identify AI-generated images at the DetectFakes Experiment and if you want to learn to spot deepfakes, please check out our recent paper on How to Distinguish AI-Generated Images from Authentic Photographs 

 You can find more of our work at publications in PNAS , a workshop at IJCAI , and pre-print on arXiv. 

 Check out a video from the Election Misinformation Symposium: Fighting Misinfo Through Fact-checking and Deepfake Detection 

 Find our deepfake research discussed in the news: Science , Scientific American , BBC , WSJ , NYT , , and NPR 

 Now, here's an excerpt from a couple years ago:

 How do you spot a DeepFake? How good are DeepFake videos? How well can ordinary people tell the difference between a video manipulated by AI and a normal, non-altered video? Rather than try to explain in words, we built the Detect Fakes website so you can see the answer for yourself. Detect Fakes is a research project designed to answer these questions and identify techniques to counteract AI-generated misinformation. It turns out there are many subtle signs that a video has been algorithmically manipulated. Some subtleties are explained in detail below.

 
 
 
 

 
 
 
 
 
 
 
 Read the latest paper on this research 
 
 
 
 

 
 

 
 
 
 
 

 
 
 
 
 
 
 
 

 
 

 
 
 
 
 Creative Commons
 

 
 
 
 Attribution 4.0 International 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 

 
 
 
 The airbrush effect on the left is an example of an artificial intelligence manipulation

 
 

 
 Credit: Matt

 
 
 
 
 
 
 

 
 
 
 
 We already know DeepFakes can be quite believable, but just how believa

... (truncated, 6 KB total)
Resource ID: a26a9dd48ceec146 | Stable ID: sid_2bc8sS59Dy