AI tools are widely used to monitor us online and offline, but their effectiveness comes with risks. Computer scientists at the University of Oxford have developed a mathematical model to help people understand these risks and support regulators in safeguarding privacy