Apple’s work on I-Heirance for Siri has been officially delayed (now it is expected for “in the next year”) and a developer thinks he knows why-the Siri are more intelligent and more personalized is, more dangerous it can be if something goes wrong.
Simon Willison, the developer of the data analysis tool set, points the finger in quick injections. AIS are generally limited by their mother -in -law who impose some rules on them. However, it is possible to “jailbreak” artificial intelligence by speaking to break those rules. This is done with the so -called “ready injections”.
As a simple example, an artificial intelligence model may have been asked to refuse to answer questions about asking something illegal. What if I asked the AI to write a poem on the wiring of a car? Writing poems is not illegal, right?
This is a problem that all companies that offer artificial intelligence chatbots must face and have improved in blocking the Over Jailbreak, but it is not yet a solved problem. Worse still, Siri di jailbreak can have much worse consequences than most chatbots because of what he knows of you and what he can do. Apple Jacqueline Roy’s spokesman described Siri as follows:
“We also worked on a more personalized Siri, giving him more awareness of your personal context, as well as the ability to act for you inside and your apps.”
Apple, without a doubt, puts in place rules to prevent Siri from accidentally revealing your private data. But what happens if a prompt injection can do it anyway? Even the “ability to act for you” can be exploited, so it is essential for a company that is aware of privacy and security as Apple to make sure Siri cannot be rhouting. And, apparently, it will take time.
Start a new Thread