Observability now equals watching AI

As enterprises embed AI deeper into core applications, system observability is no longer based on traditional metrics like uptime and latency but instead depends on factors like model accuracy, data integrity, and real-time agent behaviour. Observability now includes spotting hallucinations, detecting prompt injection and data poisoning attacks, and tracking how AI models handle personal data, said Yanbing Li, chief product officer at Datadog – a cloud monitoring and security company.

“So now you start to worry about personal information that may…