How does AI-generated disinformation work?

The spread of misinformation has reached a new dimension with artificial intelligence (AI). While AI technologies such as chatbots, image generators and language models have many positive applications, they are also misused for disinformation. In this article, we take a look at how AI can deceive - with deepfakes, social bots and fake texts. We also show why AI-generated content is so easy to manipulate.

How artificial intelligence can deceive: Deepfakes, bots & fake texts explained:

1. Deepfakes: deceptively real fake videos

Deepfakes are videos or audio files that have been manipulated using AI to make people say or do something they have never said or done. This technology is based on deep learning algorithms that can imitate faces and voices in a deceptively realistic way.

➡ Example: Politicians are made to say things in deepfake videos that they have never said in order to influence elections or social opinions.

2. Social bots: automated opinion machines

Social bots are AI-controlled programs that spread disinformation on social networks on a massive scale. They like, comment on and share posts in order to artificially reinforce certain narratives.

➡ Example: Thousands of bots specifically push fake news articles so that they appear as “trending topics” and appear more credible to real users.

3. Fake texts: AI-generated disinformation

Language models such as GPT-4 can write convincing texts - from fake news articles to invented expert analyses to opinion pieces designed to polarize.

➡ Example: An AI-generated study supposedly “proves” that a conspiracy theory is true - even though the study was completely made up.

GPT & Co.: AI-generated content is so easy to manipulate:

1. Scalability: disinformation in mass production

With AI, disinformation can be produced in gigantic quantities. Where previously individuals had to write fake news, today AI models generate thousands of texts in a matter of seconds.

2. Personalization: fake content that is tailored exactly to you

AI can adapt texts to specific target groups. This means that fake news can be individually formulated to be particularly convincing - based on the language, interests or political views of the reader.

3. Deceptively real content

Modern AI models are so good that they make real and fake content almost indistinguishable. It becomes particularly problematic when AI can also imitate images, videos and voices.

4. Lack of regulation

There are currently hardly any laws that restrict the spread of AI-generated disinformation. Many platforms do not yet have effective mechanisms to detect and stop AI fakes.

Conclusion:

Artificial intelligence makes disinformation faster, more targeted and harder to detect. Deepfakes, social bots and fake texts show just how powerful this technology can be - for both good and bad. To protect ourselves, we should all learn to critically scrutinize content, check sources and not be blinded by emotional messages. Combating AI-generated disinformation will be one of the biggest challenges of the digital future.