Hi, In the app I’m working on, we rely on SecKeychainUnlock to verify that a password can be used to unlock the login keychain. When macOS 26.4 rolled out, we started getting bug reports that led me to a discovery that makes me think SecKeychainUnlock behavior was changed. I’m going to illustrate my findings with a sample code: #include <pwd.h> #include <stdio.h> #include <string.h> #include <unistd.h> #include <Security/SecKeychain.h> #pragma clang diagnostic ignored "-Wdeprecated-declarations" int main(void) { char password[100]; printf("password: "); scanf("%s", password); struct passwd *home = getpwuid(getuid()); if (!(home && home->pw_dir)) return 1; char path[1024]; strcat(path, home->pw_dir); strcat(path, "/Library/Keychains/login.keychain-db"); SecKeychainRef keychain = NULL; OSStatus result = SecKeychainOpen(path, &keychain); if (result != errSecSuccess) { fprintf(stderr, "SecKeychainOpen failed (error %d)\n", result); return 1; } SecKeychainStatus status = 0; result = SecKeychainGetStatus(keychain, &status); if (result != errSecSuccess) { fprintf(stderr, "SecKeychainGetStatus failed (error %d)\n", result); return 1; } if (status & kSecUnlockStateStatus) { printf("keychain is unlocked, will try to lock first\n"); result = SecKeychainLock(keychain); if (result != errSecSuccess) { fprintf(stderr, "SecKeychainLock failed (error %d)\n", result); return 1; } printf("SecKeychainLock succeeded\n"); } else { printf("keychain is locked\n"); } result = SecKeychainUnlock(keychain, strlen(password), password, TRUE); if (result == errSecSuccess) { printf("SecKeychainUnlock succeeded\n"); printf("password '%s' appears to be valid\n", password); } else { printf("SecKeychainUnlock failed (error %d)\n", result); printf("password '%s' appears to be invalid\n", password); } return 0; } Here are the outputs of this program on a machine running macOS 26.3 when provided with a correct password deadbeef and with an incorrect password foobar: testuser1@tahoe1 kcdebug % ./kcdebug password: deadbeef keychain is unlocked, will try to lock first SecKeychainLock succeeded SecKeychainUnlock succeeded password 'deadbeef' appears to be valid testuser1@tahoe1 kcdebug % ./kcdebug password: foobar keychain is unlocked, will try to lock first SecKeychainLock succeeded SecKeychainUnlock failed (error -25293) password 'foobar' appears to be invalid And here are the outputs of this program on a machine running macOS 26.4: testuser1@tahoe2 kcdebug % ./kcdebug password: deadbeef keychain is unlocked, will try to lock first SecKeychainLock succeeded SecKeychainUnlock succeeded password 'deadbeef' appears to be valid testuser1@tahoe2 kcdebug % ./kcdebug password: foobar keychain is unlocked, will try to lock first SecKeychainLock succeeded SecKeychainUnlock succeeded password 'foobar' appears to be valid I’m prepared to send a feedback with Feedback Assistant, but I would like to get a confirmation that this is indeed a bug and not an intended change in behavior. I would also like to know what are my options now. SecKeychainUnlock is just a means to an end; what I really need is the ability to keep the keychain password in sync with the user password when the latter is changed by our program. Thanks in advance.

Summary I am investigating a change in macOS 26.1 (Tahoe) where plain (non-bundled) executables that request screen recording access no longer appear under: System Settings → Privacy & Security → Screen & System Audio Recording This behavior differs from macOS Sequoia, where these executables did appear in the list and could be managed through the UI. Tahoe still prompts for permission and still allows the executable to capture the screen once permission is granted, but the executable never shows up in the UI list. This breaks user expectations and removes UI-based permission management. To confirm the behavior, I created a small reproduction project with both: a plain executable, and an identical executable packaged inside an .app bundle. Only the bundled version appears in System Settings. Observed Behaviour 1. Plain Executable (from my reproduction project) When running a plain executable that captures the screen: macOS displays the normal screen-recording permission prompt. Before granting permission: screenshots show only the desktop background. After granting permission: screenshots capture the full display. The executable does not appear under “Screen & System Audio Recording”. Even when permission is granted manually (e.g., dragging the executable into the pane), the executable still does not appear, which prevents the user from modifying or revoking the permission through the UI. If the executable is launched from inside another app (e.g., VS Code, Terminal), the parent app appears in the list instead, not the executable itself. 2. Bundled App Version (from the reproduction project) I packaged the same code into a simple .app bundle (ScreenCaptureApp.app). When running the app: The same permission prompt appears. Pre-permission screenshots show the desktop background. Post-permission screenshots capture the full display. The app does appear under “Screen & System Audio Recording”. This bundle uses the same underlying executable — the only difference is packaging. Hypothesis macOS 26.1 (Tahoe) appears to require app bundles for an item to be shown in the Screen Recording privacy UI. Plain executables: still request and receive permission, still function correctly after permission is granted, but do not appear in the System Settings list. This may be an intentional change, undocumented behavior, or a regression. Reproduction Project The reproduction project includes: screen_capture.go A simple Go program that captures screenshots in a loop. screen_capture_executable Plain executable built from the Go source. ScreenCaptureApp.app/ App bundle containing the same executable. build.sh Builds both the plain executable and the app bundle. Permission reset and TCC testing scripts. The project demonstrates the behavior consistently. Steps to Reproduce Plain Executable Build: ./build.sh Reset screen capture permissions: sudo tccutil reset ScreenCapture Run: ./screen_capture_executable Before granting: screenshots show desktop only. Grant permission when prompted. After granting: full screenshots. Executable does not appear in “Screen & System Audio Recording”. Bundled App Build (if not already built): ./build.sh Reset permissions (optional): sudo tccutil reset ScreenCapture Run: open ScreenCaptureApp.app Before granting: screenshots show desktop. After granting: full screenshots. App bundle appears in the System Settings list. Additional Check I also tested launching the plain executable as a child process of another executable, similar to how some software architectures work. Result: Permission prompt appears Permission can be granted Executable still does not appear in the UI, even though TCC tracks it internally → consistent with the plain-executable behaviour. This reinforces that only app bundles are listed. Questions for Apple Is the removal of plain executables from “Screen & System Audio Recording” an intentional change in macOS Tahoe? If so, does Apple now require all screen-recording capable binaries to be packaged as .app bundles for the UI to display them? Is there a supported method for making a plain executable (launched by a parent process) appear in the list? If this is not intentional, what is the recommended path for reporting this as a regression? Files Unfortunately, I have discovered the zip file that contains my reproduction project can't be directly uploaded here. Here is a Google Drive link instead: https://drive.google.com/file/d/1sXsr3Q0g6_UzlOIL54P5wbS7yBkpMJ7A/view?usp=sharing Thank you for taking the time to review this. Any insight into whether this change is intentional or a regression would be very helpful.

QuickLookAR shares the actual USDZ model instead of the original website URL — critical copyright and data leak issue on iOS 26 Since iOS 26, QuickLookAR (or ARQuickLookPreviewItem) no longer preserves the original web URL when sharing a model. Instead of sending the link to the hosted file, the system directly shares the actual USDZ model file with the recipient. This is a critical regression and a severe breach of intellectual property protection, as it exposes proprietary 3D models that must never be distributed outside of the controlled web environment. In earlier iOS versions (tested up to iOS 18), QuickLookAR correctly handled sharing — the share sheet would send the website link where the model is hosted, not the file itself. Starting with iOS 26, this behavior has changed and completely breaks the intended secure flow for AR experiences. Our project relies on allowing users to view models in AR via QuickLook, without ever transferring the underlying 3D assets. Now, the share operation forces full file sharing, giving end users unrestricted access to the model file, which can be copied, rehosted, or reverse-engineered. This issue critically affects production environments and prevents us from deploying our AR-based solutions. Implement a standard QuickLookAR preview with a USDZ file hosted on your web server (e.g., via ARQuickLookPreviewItem). 2. Open the AR view on iOS 26. 3. Tap the Share icon from QuickLookAR. 4. Send via any messenger (Telegram, WhatsApp, etc.). 5. Observe that the actual .usdz model is sent instead of the original website URL. ⸻ Expected behavior: QuickLookAR should share only the original URL (as in iOS 17–18), not the file itself. This ensures that intellectual property and licensed 3D models remain protected and controlled by the content owner. ⸻ Actual behavior: QuickLookAR shares the entire USDZ file, leaking the model content outside of the intended environment. ⸻ Impact: • Violation of copyright and confidential data policies • Loss of control over proprietary 3D assets • Breaking change for all existing web-based AR integrations • Critical blocker for AR production deployment ⸻ Environment: • iOS 26.0 and 26.1 (tested on iPhone 14, iPhone 15) • Safari + QuickLookAR integration • Works correctly on iOS 17 / iOS 18 ⸻ Notes: This regression appears to have been introduced in the latest iOS 26 system handling of QuickLookAR sharing. Please escalate this issue to the ARKit / QuickLook engineering team as it directly affects compliance, IP protection, and usability of AR features across production applications. Additional Notes / Verification: Please test this behavior yourself using the CheckAR test model on my website: https://admixreality.com/ios26/ • If the login page appears, click “Check AR” and then “View in Your Space”. • On iOS 18 and earlier, sharing correctly sends the website URL. • On iOS 26, sharing sends the actual USDZ model file. This clearly demonstrates the regression and the security/IP issue.

Hi Apple, Currently we want to have enhancement for SMS OTP that we want to implement OTP Autofill, But after do some research we're stuck with option that the OTP only show in keyboard suggestion, is there any way for making OTP is automatically filled without user have to click the keyboard suggestion when receiving the SMS. Thanks Best Regards, Admiral Sultano Harly.

I'm trying to export and re-import a P-256 private key that was originally generated via SecKeyCreateRandomKey(), but I keep running into roadblocks. The key is simply exported via SecItemExport() with format formatWrappedPKCS8, and I did set a password just to be sure. Do note that I must use the file-based keychain, as the data protection keychain requires a restricted entitlement and I'm not going to pay a yearly fee just to securely store some private keys for a personal project. The 7-day limit for unsigned/self-signed binaries isn't feasible either. Here's pretty much everything I could think of trying: Simply using SecItemImport() does import the key, but I cannot set kSecAttrLabel and more importantly: kSecAttrApplicationTag. There just isn't any way to pass these attributes upfront, so it's always imported as Imported Private Key with an empty comment. Keys don't support many attributes to begin with and I need something that's unique to my program but shared across all the relevant key entries, otherwise it's impossible to query for only my program's keys. kSecAttrLabel is already used for something else and is always unique, which really only leaves kSecAttrApplicationTag. I've already accepted that this can be changed via Keychain Access, as this attribute should end up as the entry's comment. At least, that's how it works with SecKeyCreateRandomKey() and SecItemCopyMatching(). I'm trying to get that same behaviour for imports. Running SecItemUpdate() afterwards to set these 2 attributes doesn't work either, as now the kSecAttrApplicationTag is suddenly used for the entry's label instead of the comment. Even setting kSecAttrComment (just to be certain) doesn't change the comment. I think kSecAttrApplicationTag might be a creation-time attribute only, and since SecItemImport() already created a SecKey I will never be able to set this. It likely falls back to updating the label because it needs to target something that is still mutable? Using SecItemImport() with a nil keychain (i.e. create a transient key), then persisting that with SecItemAdd() via kSecValueRef does allow me to set the 2 attributes, but now the ACL is lost. Or more precise: the ACL does seem to exist as any OS prompts do show the label I originally set for the ACL, but in Keychain Access it shows as Allow all applications to access this item. I'm looking to enable Confirm before allowing access and add my own program to the Always allow access by these applications list. Private keys outright being open to all programs is of course not acceptable, and I can indeed access them from other programs without any prompts. Changing the ACL via SecKeychainItemSetAccess() after SecItemAdd() doesn't seem to do anything. It apparently succeeds but nothing changes. I also reopened Keychain Access to make sure it's not a UI "caching" issue. Creating a transient key first, then getting the raw key via SecKeyCopyExternalRepresentation() and passing that to SecItemAdd() via kSecValueData results in The specified attribute does not exist. This error only disappears if I remove almost all of the attributes. I can pass only kSecValueData, kSecClass and kSecAttrApplicationTag, but then I get The specified item already exists in the keychain errors. I found a doc that explains what determines uniqueness, so here are the rest of the attributes I'm using for SecItemAdd(): kSecClass: not mentioned as part of the primary key but still required, otherwise you'll get One or more parameters passed to a function were not valid. kSecAttrLabel: needed for my use case and not part of the primary key either, but as I said this results in The specified attribute does not exist. kSecAttrApplicationLabel: The specified attribute does not exist. As I understand it this should be the SHA1 hash of the public key, passed as Data. Just omitting it would certainly be an option if the other attributes actually worked, but right now I'm passing it to try and construct a truly unique primary key. kSecAttrApplicationTag: The specified item already exists in the keychain. kSecAttrKeySizeInBits: The specified attribute does not exist. kSecAttrEffectiveKeySize: The specified attribute does not exist. kSecAttrKeyClass: The specified attribute does not exist. kSecAttrKeyType: The specified attribute does not exist. It looks like only kSecAttrApplicationTag is accepted, but still ignored for the primary key. Even entering something that is guaranteed to be unique still results in The specified item already exists in the keychain, so I think might actually be targeting literally any key. I decided to create a completely new keychain and import it there (which does succeed), but the key is completely broken. There's no Kind and Usage at the top of Keychain Access and the table view just below it shows symmetric key instead of private. The kSecAttrApplicationTag I'm passing is still being used as the label instead of the comment and there's no ACL. I can't even delete this key because Keychain Access complains that A missing value was detected. It seems like the key doesn't really contain anything unique for its primary key, so it will always match any existing key. Using SecKeyCreateWithData() and then using that key as the kSecValueRef for SecItemAdd() results in A required entitlement isn't present. I also have to add kSecUseDataProtectionKeychain: false to SecItemAdd() (even though that should already be the default) but then I get The specified item is no longer valid. It may have been deleted from the keychain. This occurs even if I decrypt the PKCS8 manually instead of via SecItemImport(), so it's at least not like it's detecting the transient key somehow. No combination of kSecAttrIsPermanent, kSecUseDataProtectionKeychain and kSecUseKeychain on either SecKeyCreateWithData() or SecItemAdd() changes anything. I also tried PKCS12 despite that it always expects an "identity" (key + cert), while I only have (and need) a private key. Exporting as formatPKCS12 and importing it with itemTypeAggregate (or itemTypeUnknown) does import the key, and now it's only missing the kSecAttrApplicationTag as the original label is automatically included in the PKCS12. The outItems parameter contains an empty list though, which sort of makes sense because I'm not importing a full "identity". I can at least target the key by kSecAttrLabel for SecItemUpdate(), but any attempt to update the comment once again changes the label so it's not really any better than before. SecPKCS12Import() doesn't even import anything at all, even though it does return errSecSuccess while also passing kSecImportExportKeychain explicitly. Is there literally no way?

The One-time codes documentation details how to enable autofill for SMS based codes. However, there is no details about how to correctly implement autofill for email based codes. I am observing the email based autofill works inconsistently when using email based OTC. In my application: There is latency of 10-15 seconds from when the email arrives to when it is available for autofill. After the autofill feature is used, the OTC email is not being deleted from the inbox automatically. Without documentation, it's unclear to me what I might be doing wrong that is causing these side effects. I found an ietf proposal for how autofill with email based codes might work, but it’s unclear if this is how Apple has implemented the feature: https://www.ietf.org/archive/id/draft-wells-origin-bound-one-time-codes-00.html#name-email Existing docs for Autofill using SMS: https://developer.apple.com/documentation/security/enabling-autofill-for-domain-bound-sms-codes

I have a binary executable which needs to be given Accessibility Permissions so it can inject keypresses and mouse moves. This was always possible up to macOS 15 - when the first keypress arrived the Accessibility Permissions window would open and allow me to add the executable. However this no longer works in macOS 26: the window still opens, I navigate to the executable file and select it but it doesn't appear in the list. No error message appears. I'm guessing that this may be due to some tightening of security in Tahoe but I need to figure out what to change with my executable to allow it to work.

Hi! Is it possible to disable the option for users to 'Sign in with Another Device'? I encounter this message during the authentication process and I want to prevent it from appearing. I appreciate your help and look forward to your response.

We have successfully deployed our Qt C++ application on Windows and Android using OpenSSL with TLS Pre-Shared Key (PSK) authentication to connect to our servers. However, I understand that apps submitted to the App Store must use SecureTransport as the TLS backend on iOS. My understandiunig is that SecureTransport does not support PSK ciphersuites, which is critical for our security architecture. Questions: Does SecureTransport support TLS PSK authentication, or are there plans to add this feature? If PSK is not supported, what is Apple's recommended alternative for applications that require PSK-based authentication? Is there an approved exception process that would allow me to use OpenSSL for TLS connections on iOS while still complying with App Store guidelines? The application requires PSK for secure communication with our infrastructure, and we need guidance on how to maintain feature parity across all platforms while meeting App Store requirements

Hello, sorry for the awkward text formatting but I kept getting prevented from positing due to "sensitive language"... Help.txt

With the ScreenTime API Apple talks a lot about their focus on privacy and the data not leaving the device. Does that mean there would be a problem with an app where the users ScreenTime data is shared to a custom backend? Could this potentially cause an app to be rejected from the AppStore?

Hi, We are implementing Platform SSO and using attestKey during registration via ASAuthorizationProviderExtensionLoginManager. Could you clarify whether the attestKey flow involves sending attestation data to an Apple server for verification (similar to App Attest in the DeviceCheck framework), or if the attestation certificate chain is generated and signed entirely on-device without any Apple server interaction? The App Attest flow is clearly documented as using Apple’s attestation service, but the Platform SSO process is less clearly described. Thank you.

I have a project with a single app target that serves two environments, and two schemes, one for each env, using xcconfig files for defining environment-specific stuff. I'm trying to figure this out for months, so I've tried multiple approaches throughout this period: Have a single domain in "Associated domains" in Xcode, defined as webcredentials:X where X gets replaced using a value from xcconfig. Have two domain entries in "Associated domains" webcredentials:PROD_DOMAIN and webcredentials:STAGING_DOMAIN. Have a different order of domains Results are very interesting: whatever I do, whatever approach I take, password autofill works on staging, but doesn't work on production. I'm aware that we need to test production on Test Flight and AppStore builds. That's how we're testing it, and it's not working. Tested on multiple devices, on multiple networks (wifi + mobile data), in multiple countries.. you name it. The server side team has checked their implementation a dozen times; it's all configured properly, in the exact same way across environments (except bundle ID, ofc). We tried a couple websites for validating the apple-app-site-association file, and while all of those are focused on testing universal links, they all reported that the file is configured properly. Still, password autofill doesn't work. I prefer not to share my app's domains publicly here. Ideally I would contact Apple Developer Support directly, but they now require a test project for that, and since 'a test project' is not applicable to my issue, I'm posting here instead.

Hi, I’m building a macOS tool that analyzes process behavior to detect autonomous / AI-like activity locally (process trees, file access patterns, and network usage). The system is fully user-space and runs locally in real time. I’m planning to use the Endpoint Security Framework for process and file event monitoring. This is an open-source project (non-enterprise), developed by a solo developer. My question: What are the realistic chances of getting Endpoint Security entitlements approved for this type of project? Are there specific requirements or common reasons for rejection I should be aware of? Thanks, sivan-rnd

In these threads, it was clarified that Credential Provider Extensions must set both Backup Eligible (BE) and Backup State (BS) flags to 1 in authenticator data: https://developer.apple.com/forums/thread/745605 https://developer.apple.com/forums/thread/787629 However, I'm developing a passkey manager that intentionally stores credentials only on the local device. My implementation uses: kSecAttrAccessibleWhenUnlockedThisDeviceOnly for keychain items kSecAttrTokenIDSecureEnclave for private keys No iCloud sync or backup These credentials are, by definition, single-device credentials. According to the WebAuthn specification, they should be represented with BE=0, BS=0. Currently, I'm forced to set BE=1, BS=1 to make the extension work, which misrepresents the actual backup status to relying parties. This is problematic because: Servers using BE/BS flags for security policies will incorrectly classify these as synced passkeys Users who specifically want device-bound credentials for higher security cannot get accurate flag representation Request: Please allow Credential Provider Extensions to return credentials with BE=0, BS=0 for legitimate device-bound passkey implementations. Environment: macOS 26.2 (25C56), Xcode 26.2 (17C52)

Hey all, Question for the masses.... Does the Yubikey authentication have a OS dependency and it only works with a stable, public OS? Does Azure/Okta/Yubikey beta OS26? My CEO installed iPadOS 26 on his iPad and was not able to authenticate via Yubikey into our company environment. I ran the same scenario on my iPad using iPadOS 26 and I had the same results. Downgrading to iPAdOS doesn't pose these issues. I'm assuming something isn't fine-tuned yet?

Hello, I am using the prf extension for passkeys that is available since ios 18 and macos15. I am using a fixed, hardcoded prf input when creating or geting the credentials. After creating a passkey, i try to get the credentials and retrieve the prf output, which works great, but i am getting different prf outputs for the same credential and same prf input used in the following scenarios: Logging in directly (platform authenticator) on my macbook/iphone/ipad i get "prf output X" consistently for the 3 devices When i use my iphone/ipad to scan the qr code on my macbook (cross-platform authenticator) i get "prf output Y" consistently with both my ipad and iphone. Is this intended? Is there a way to get deterministic prf output for both platform and cross-platform auth attachements while using the same credential and prf input?

I am trying to setup remote Java debugging between two machines running macOS (15.6 and 26). I am able to get the Java program to listen on a socket. However, I can connect to that socket only from the same machine, not from another machine on my local network. I use nc to test the connection. It reports Connection refused when trying to connect from the other machine. This issue sounds like it could be caused by the Java program lacking Local Network system permission. I am familiar with that issue arising when a program attempts to connect to a port on the local network. In that case, a dialog is displayed and System Settings can be used to grant Local Network permission to the client program. I don't know whether the same permission is required on the program that is receiving client requests. If it is, then I don't know how to grant that permission. There is no dialog, and System Settings does not provide any obvious way to grant permission to a program that I specify. Note that a Java application is a program run by the java command, not a bundled application. The java command contains a hard-wired Info.plist which, annoyingly, requests permission to use the microphone, but not Local Network access.

Hello, I’m storing some values in the Keychain with the attribute ‘ksecattraccessibleafterfirstunlockthisdeviceonly’ (https://developer.apple.com/documentation/security/ksecattraccessibleafterfirstunlockthisdeviceonly). When I migrate user data between iPhones via iCloud, this behaves as expected and the keys are not preserved. However, when I migrate using a direct connection between two devices, the keys are preserved, which seems to contradict the attribute’s intent. Is this a known behavior, and if so, is there a workaround?

We have a crash on DCDevice.current.isSupported We want to try to make a serial queue to generate tokens but the side effect would be the same token would be used on multiple server API requests that are made within a few ms of each other? Is this safe or will the Apple server immediately reject the same token being reused? Can you share how long tokens are safe to use for? Here is the code we want to try final actor DeviceTokenController: NSObject { static var shared: DeviceTokenController = .init() private var tokenGenerationTask: Task<Data?, Never>? var ephemeralDeviceToken: Data? { get async { // Re-using the token for short periods of time if let existingTask = tokenGenerationTask { return await existingTask.value } let task = Task<Data?, Never> { guard DCDevice.current.isSupported else { return nil } do { return try await DCDevice.current.generateToken() } catch { Log("Failed to generate ephemeral device token", error) return nil } } tokenGenerationTask = task let result = await task.value tokenGenerationTask = nil return result } } }