Deepfakes and Disinformation: AI Misuse Threatens Democracy, Study Reveals

A new study from Google’s DeepMind division sheds light on the most common malicious uses of AI, revealing that political deepfakes and disinformation campaigns are the top concerns.

Key Findings: Deepfakes and Disinformation Dominate AI Misuse; The study, conducted in collaboration with Google’s Jigsaw unit, analyzed around 200 incidents of AI misuse and found that:

• The creation of realistic but fake images, videos, and audio of people, known as deepfakes, was the most prevalent form of AI misuse, nearly twice as common as the next highest category.
• The second most common misuse was the falsification of information using text-based tools like chatbots to generate misinformation for online posting.
• Shaping public opinion was the primary goal behind 27% of the misuse cases, raising concerns about the potential impact on elections worldwide.

Implications for Democracy: Distorting Sociopolitical Reality; The findings highlight the risks posed by AI-generated content to the democratic process:

• Deepfakes of political leaders, such as UK Prime Minister Rishi Sunak, have appeared on various social media platforms in recent months, coinciding with the UK’s upcoming general election.
• Ardi Janjeva from The Alan Turing Institute emphasized the long-term risks to democracies posed by the distortion of collective understanding of sociopolitical reality through the contamination of public information with AI-generated content.

Motivations and Accessibility: Money and Ease of Use; The study revealed insights into the drivers behind AI misuse and the low barriers to entry:

• Making money was the second most common motivation, with bad actors offering services to create deepfakes, including generating fake nude images of real people, or using AI to produce large volumes of content like fake news articles.
• Most incidents involved easily accessible tools that required minimal technical expertise, enabling a wider range of bad actors to misuse generative AI.

Industry Response: Monitoring and Mitigation; As generative AI products gain popularity, companies are starting to monitor the flood of misinformation and harmful content created by their tools:

• OpenAI, the creator of ChatGPT, released research in May exposing state-linked operations from Russia, China, Iran, and Israel that used its tools to create and spread disinformation.
• Google DeepMind’s research aims to influence how the company improves its safety evaluations for AI models and hopes to shape the perspectives of competitors and stakeholders on how AI-related harms manifest.

Looking Ahead: Addressing the Challenges; The DeepMind study underscores the urgent need for proactive measures to counter the malicious use of AI, particularly in the context of safeguarding democratic processes. As AI technologies continue to advance and become more accessible, collaboration between researchers, industry leaders, and policymakers will be crucial in developing effective strategies to mitigate the risks posed by deepfakes and AI-generated disinformation.