Top News

Apple has announced one of the most practically useful AI features in iOS 27, an intelligent agent built directly into the Passwords app that will automatically visit websites and change your insecure or compromised passwords without requiring you to do it manually.

The feature uses Apple Intelligence and Safari working together to act agentically on the user’s behalf. In plain terms, the Passwords app identifies which of your saved passwords are weak, reused, or flagged as compromised, and then Siri AI goes to each individual website, logs in, navigates to the password change page, generates a strong new password, and updates it, all without the user having to open a single browser tab.

Why this is a bigger deal than it sounds

Most people know they have weak or reused passwords. The reason they do not fix them is not ignorance but friction. Changing a password on even one website requires opening the site, finding the account settings, navigating to the security section, entering the old password, generating or typing a new one, and saving it. Multiply that by the dozens of accounts a typical user has flagged for improvement and the task becomes something that gets indefinitely postponed.

Apple’s password agent eliminates that friction entirely. The AI handles the navigation, the generation, and the updating. The user receives a notification that their passwords have been secured.

What it means competitively

Google Password Manager and LastPass have offered insecure password identification for years, but neither has automated the remediation step to this degree. Apple is going a step further by not just telling you what is wrong but fixing it for you, which is the logical end point of what a genuinely useful AI agent in a password manager should do.

It is also Apple’s clearest signal yet that agentic AI, where software acts independently in the background to complete multi-step tasks, is a core part of where iOS is heading. The password agent is a relatively low-stakes and high-trust introduction to that paradigm, a category where users are likely comfortable letting an AI act autonomously because the downside risk of the agent making an error is low and the benefit is concrete and immediate.