AI also steps in to remember passwords

As a new step in making our daily online activities easier through Artificial Intelligence (AI), the ability to manage passwords and access various accounts has now been introduced. Through this, AI agents like 'Claude' from 'Anthropic' can automatically perform various tasks, from purchasing groceries to checking retirement funds.

To perform these tasks securely, '1Password', a well-known password management software, has introduced a special method. This allows the AI agent temporary access to relevant websites without exposing the user's actual passwords to the AI model or the respective company. In this process, the AI can only see the username and the relevant website, and the password is never visible to it.

Several steps are implemented here to ensure security. Every time an AI agent attempts to access an account, biometric authentication such as the user's fingerprint (Touch ID) is mandatory, and the granted permission is valid only for that specific task. Furthermore, measures have been taken to lock and secure all other data in 1Password when the AI agent is active.

However, experts in the field point out that certain cybersecurity risks may arise when using this technology. Specifically, through cyberattacks like "prompt injection," computer hackers can trick AI using things like calendar invites to give secret instructions to change master passwords or steal data. It has also been further revealed that AI agents browsing the internet are at risk of falling victim to such hidden malicious instructions through comments and reviews on various websites.