Apple’s upcoming iOS 27, iPadOS 27, and macOS 27 will feature an agentic tool that automatically updates compromised website passwords, raising significant questions regarding security, authority, and user control.

Key Points

• The feature uses Apple Intelligence and Safari to navigate websites, replace weak passwords, and save new credentials as a Live Activity.
• Security experts warn that granting an AI agent authority to modify account credentials creates potential risks, including prompt injection and account lockouts.
• The system is currently in developer beta as of June 8, 2026, with specific security architecture and failure-handling protocols yet to be fully documented.
• Recommended safeguards include isolating credentials from the AI model, requiring biometric user approval for changes, and limiting the agent to password-only modifications.
• The tool aims to address the common security failure where users ignore breach notifications or fail to update compromised credentials effectively.

Why it Matters

Automating high-impact actions like password rotation shifts the security burden from the user to an autonomous agent, which could significantly improve credential hygiene if implemented correctly. However, because these agents operate within the untrusted environment of the open web, they must be strictly constrained to prevent malicious exploitation or accidental account lockouts.