Apple may have delayed the Siri upgrade for fear of jailbreaks | Infinium-tech
Apple’s work on AI-Enhancement for Siri is officially delayed (now it is “slaughtered to roll out” in the coming year) and a developer feels that they know why they know why and more individual Siri is, the more dangerous it can be if something is wrong.
The developer of the data analysis tool dataset, Simon Villison, indicates the finger on early injections. AIS are generally banned by their original companies that apply certain rules to them. However, it is possible to “jailbreak” AI by talking in breaking those rules. This is done with the so -called “quick injection”.
As a simple example, an AI model may be instructed to refuse to answer questions about doing something illegal. But what if you ask AI to write a poem to a car about hotwiring? Writing poems is not illegal, right?
This is an issue that all companies offer AI chatbots and they have become better in blocking clear gelbreaks, but this is not yet a problem solved. Worse than that of most of the chatbots of gelbracing Siri can be much worse results because what it knows about you and what it can do. Apple spokesman Jacqueline Roy describes Siri as follows:
“We are also working on a more individual Siri, giving it more awareness about your personal reference, as well as the ability to take action within your apps and for you.”
Apple, undoubtedly, put the rules to prevent the Siri from revealing your personal data. But what if a quick injection can be found to do it anyway? “Your ability to take action” can also be exploited, so it is important for a company that is alert to privacy and security as Apple to ensure that Siri cannot be jailed. And, of course, it is going to take some time.
Leave a Reply