What is PasswordBreachAgent on Mac?

Last updated: 2026-07-31

What it is

PasswordBreachAgent is an Apple-supplied background process that runs separately for each signed-in user. It supports the “Detect Compromised Passwords” feature used by Passwords and Safari. Its job is to compare passwords saved for AutoFill with a curated set of password values found in known data leaks.

The comparison has two stages. Commonly leaked passwords can be matched against a list stored locally on the Mac. Passwords not covered by that local list are checked through a cryptographic private-set-intersection protocol involving Apple servers. This design lets the Mac determine whether a password appears in the breach set without sending Apple the actual password or the associated account identifier.

That privacy protection does not mean that every part of the check stays on the Mac. For a remote comparison, PasswordBreachAgent transmits calculated password-derived information. Apple documents this as including a 15-bit prefix derived from SHA-256 and a cryptographically blinded elliptic-curve point.

PasswordBreachAgent is part of macOS, not an application installed by an unknown third party. The executable is signed as an Apple platform component, managed by launchd, and delivered inside the macOS system cryptex.

Why it talks to the network

Network activity begins when Detect Compromised Passwords is enabled. The Mac performs round-robin checks periodically, following a background schedule that does not depend on whether the user has recently opened Passwords or used the password manager. It may also receive an updated local list of commonly leaked passwords.

A password found in the local list does not require the remote comparison. For a password outside that list, PasswordBreachAgent sends Apple a blinded elliptic-curve point and a bucket selector based on a 15-bit SHA-256 prefix. It receives the corresponding breach-set bucket and the server’s response, which allows the private-set-intersection check to be completed.

These connections go to Apple Password Monitoring services. PasswordBreachAgent does not contact every website for which a password has been saved, and its network activity is not a replay of Safari browsing.

In the installed launchd job examined during the research, the activity was declared as a repeating, bidirectional, network-required PeriodicLookupActivity with a one-week interval and scheduling grace. That observation describes the examined system configuration, not a permanent macOS contract. Apple does not promise that the interval will remain the same across releases, and the available evidence did not verify whether additional on-demand triggers exist.

How much traffic is normal

There is no dependable public byte range for PasswordBreachAgent. Apple has not published a stable minimum, maximum, or typical transfer size, and the available local evidence did not establish one.

Its activity can range from comparisons completed entirely against the local list to intermittent, batched, bidirectional transfers for private-set-intersection checks and list updates. It is periodic background activity rather than continuous, media-like streaming. That qualitative description does not justify converting the traffic into a specific number.

Claims that PasswordBreachAgent always uses only a few kilobytes, only a few megabytes, or less data than a web page are unsupported. A measurement from one Mac or one observation window should not be treated as a universal baseline, especially because some checks can finish locally while others need a server exchange.

Can you turn it off

Yes. The documented control is located at:

Passwords app → Passwords menu → Settings… → Detect Compromised Passwords

The preferable default is to keep the feature enabled. Disabling it removes a meaningful security warning, while no verified evidence establishes a predictable traffic saving.

Turning it off stops ongoing comparisons with known-leak data and prevents future compromised-password alerts. It does not delete passwords already saved on the Mac, and it does not disable Password AutoFill. Recommendations generated locally for weak or reused passwords may still appear because they are not necessarily dependent on the breach-monitoring service.

Force-quitting PasswordBreachAgent is not equivalent to changing this setting: launchd can start the process again. Deleting or modifying the protected system executable is unsupported. If disabling the monitoring is an intentional tradeoff, the Passwords setting is the documented control.

What people get wrong

• “PasswordBreachAgent is malware, spyware, or a keylogger.” It is an Apple platform-signed system executable, managed by launchd and shipped in the macOS system cryptex. Seeing an unfamiliar process name or a network connection does not turn it into third-party malware.

• “It uploads my actual passwords or the whole keychain.” Apple says actual passwords and account identifiers are not shared with Apple. However, saying that nothing leaves the Mac is also inaccurate: remote comparisons send calculated information, including a 15-bit hash prefix and a cryptographically blinded value.

• “A warning means this Mac or that exact account was hacked.” The documented protocol checks whether a password value appears in known-leak data. A match does not identify the organization that leaked it, prove that the saved account was the source, or establish that someone accessed that particular account.

• “This is Safari browsing traffic or always-on telemetry.” PasswordBreachAgent is a separate background agent for password monitoring. The verified evidence establishes scheduled periodic work, while any additional on-demand triggers remain unverified. The Safari-prefixed identifier is historical or internal naming; it does not mean the agent records browsing.

• “Force-quitting or deleting it is a harmless network optimization.” launchd can start it again, and changing protected macOS files is unsupported. More importantly, disabling the underlying feature removes future compromised-password warnings without any verified guarantee of a meaningful traffic reduction. The documented Passwords setting is the appropriate control when that tradeoff is intentional.

• “Its traffic is known to be a particular small number of KB or MB.” No dependable figure was verified, and Apple publishes no stable byte range. Local-only checks, remote comparisons, batching, and list updates make a universal fixed estimate unjustified.

Seeing what it actually used

To answer the question for your own Mac, open Bytetally’s per-process statistics and find PasswordBreachAgent in the process list. Because its work is periodic and some comparisons stay local, a short window showing zero bytes does not prove that it never uses the network. Check a representative longer window and treat the measured upload and download totals as evidence for that Mac, not as a universal baseline.

Related processes

Common questions