What’s particularly concerning isn’t that AI was used to carry out a cyberattack, but that it knew what to do when its first attempt failed. When it hit a security barrier, it didn’t give up or wait for new instructions. It changed tactics and looked for another way in.

We’re moving towards attacks that can think on their feet, adapt to defences and keep probing for weaknesses without a human directing every move. With millions of sensitive patient records potentially exposed in this attack, the consequences extend far beyond compromised systems, putting individuals at risk of identity theft, extortion and highly targeted scams.

As AI becomes increasingly capable of acting independently, businesses need to think carefully about how much autonomy they’re giving these systems. AI agents are rapidly becoming part of the digital workforce, but they cannot be given free rein. Organisations need to apply the same scrutiny to their digital workers as they do to their human employees, with clear permissions, oversight and the ability to intervene.