"We knew that we wanted to come up with some sort of prompt that would evade the pre-filtering, the post-filtering, as well as any guardrails within the model itself, so we started probing the model."
"The larger security issue that is prompt injection remains 'a cat and mouse problem.' Models will become better and better at identifying."
Apple Intelligence, integrated into newer Apple devices, is vulnerable to prompt injection attacks that can manipulate its output. Researchers at RSAC demonstrated a 76% success rate in bypassing security measures using two techniques. They disclosed their findings to Apple, which subsequently released updates in iOS 26.4 and macOS 26.4 to address the vulnerabilities. Despite the fix, the broader issue of prompt injection remains a persistent challenge in AI security, as models continue to evolve in their ability to detect such attacks.
Read at Theregister
Unable to calculate read time
Collection
[
|
...
]