AI Models Can Identify Keystrokes from Audio with High Accuracy, Raising Privacy Concerns

Researchers have developed AI models that can identify keystrokes by analyzing audio recordings with over 90% accuracy.
The models can discern keystrokes from recordings made via computer microphones, smartphones, or apps like Zoom.
Experts warn these "acoustic attacks" could allow hackers to steal passwords, data, and other sensitive information.
Possible defenses include randomized passwords, adding fake keystrokes, changing typing habits, and relying on biometric authentication.
The research highlights growing concerns over audio recording capabilities of devices and the privacy risks of AI.