AI safety and security
Also: KI-Security
Measures that prevent AI systems from causing harm – through errors, misuse or attacks.
The term covers two sides: “safety” means a system behaves safely even in unexpected situations and cannot be misused for dangerous purposes. “Security” means protection against attacks such as prompt injection, data theft or manipulated training data.
For businesses this means limiting the permissions of AI systems, checking inputs and outputs, logging activity and carrying out security tests before deployment.
General information, not legal or professional advice.