Prompt Injection Vulnerability Exposes Apple Intelligence to Attacks

• •Apple Intelligence is vulnerable to prompt injection attacks, affecting millions of devices.
• •RSAC researchers achieved a 76% success rate in exploiting the vulnerability using advanced techniques.
• •Apple has released patches to mitigate the issue, but prompt injection remains a persistent challenge.

Research from RSAC has revealed that Apple's on-device AI, Apple Intelligence, is vulnerable to prompt injection attacks, allowing attackers to manipulate the model's output. The researchers achieved a 76% success rate in 100 tests, using techniques such as Neural Exec and Unicode obfuscation to bypass security measures. This vulnerability affects millions of users, with estimates of 200 million Apple Intelligence-capable devices and up to 1 million apps potentially at risk. The attack can lead to the generation of offensive content or the exposure of sensitive data through connected applications. Apple was informed of these findings on October 15, 2025, and has since released updates in iOS 26.4 and macOS 26.4 to address the issue. However, the broader challenge of prompt injection remains a significant concern in AI security. The integration of Apple Intelligence into various applications increases the potential attack surface, raising the stakes for user privacy and security.