Adversarial examples are a term from the fields of Artificial Intelligence, cybercrime and cybersecurity, and digital transformation. They describe a special type of data that has been deliberately altered to intentionally mislead machines such as Artificial Intelligence (AI).

Imagine a camera automatically checking whether a stop sign is visible in a photo. An "adversarial example" would be an image that clearly shows a stop sign to our human eyes. However, if the image is deliberately manipulated with small, usually invisible changes, the AI can suddenly no longer recognise that it is a stop sign – even though a human would not notice the difference.

This phenomenon is important for companies to be aware of because hackers can use adversarial examples to trick facial recognition systems or other automated checks. Therefore, security and protection against such attacks are particularly relevant when developing AI systems. It is worth considering not only the functionality when deploying artificial intelligence, but also the security against adversarial examples.

Adversarial examples are a term from the fields of Artificial Intelligence, cybercrime and cybersecurity, and digital transformation. They describe a special type of data that has been deliberately altered to intentionally mislead machines such as Artificial Intelligence (AI).

Imagine a camera automatically checking whether a stop sign is visible in a photo. An "adversarial example" would be an image that clearly shows a stop sign to our human eyes. However, if the image is deliberately manipulated with small, usually invisible changes, the AI can suddenly no longer recognise that it is a stop sign – even though a human would not notice the difference.

This phenomenon is important for companies to be aware of because hackers can use adversarial examples to trick facial recognition systems or other automated checks. Therefore, security and protection against such attacks are particularly relevant when developing AI systems. It is worth considering not only the functionality when deploying artificial intelligence, but also the security against adversarial examples.