Fraud defences on a phone tend to fire after money has already moved. A toggle added to the newest iPhone software tries an earlier moment, weighing whether a conversation is being steered by a criminal while it is still underway.
The assessment is built from signals about the handset and the account signed into it. What leaves the device is a single word, a risk level, and nothing else. The data behind that verdict stays behind.
Scoring a scam while it happens
That split matters. An app handed a high score can slow a transaction, demand fresh verification or raise a warning. Apple’s existing filters cannot catch a coaching con, because the owner performs the action willingly. A score gives the app a reason to interrupt first.
The setting is called Impersonation Risk Detection, and it arrives switched off. Enabling it takes a trip to Privacy and Security in Settings, where a Share with App Developers toggle waits. Only participating apps can request the signal, and Apple has not named them.
Where the protection stops
Underneath sits a service Apple calls Trust Insights, which returns the verdict while stripping the inputs used to calculate it. A score is advisory, though. An app can ignore it, and the toggle only governs whether the assessment leaves the device at all.
Closing the gap between a warning and a blocked transfer remains the developer’s job, so the practical benefit depends on how many banks and payment apps adopt the setting.
