Skip to content
Browse

Edition

Sections

Appearance

AboutSourcesContactPrivacy

AI models will change their response if you warn that it’s being monitored, but not necessarily for good

A new study involving 10 AI models suggests that their willingness to follow instructions can change depending on whether they believe their responses are being monitored. The findings raise questions about how reliably AI systems behave outside controlled testing environments.

First reported by Digital Trends 1 source
Image: Digital Trends
Image: Digital Trends
Read at Digital Trends(opens in a new tab)

Full coverage

Every report we found on this story, oldest first. Each links to the original publisher.

  1. Digital Trends

    AI models will change their response if you warn that it’s being monitored, but not necessarily for good (opens Digital Trends in a new tab)

    A new study involving 10 AI models suggests that their willingness to follow instructions can change depending on whether they believe their responses are being monitored. The findings raise questions about how reliably AI systems behave outside controlled…

Something wrong with how we've grouped or summarised this story? Report a problem