Apple announced that it has postponed artificial intelligence -oriented improvements for Siri and that this update will be released next year. A software developer thinks that there may be safety concerns behind this delay. Siri’s smarter and more personalized can cause serious security problems in a wrong use scenario.
Simon Willon, the developer of the data analysis tool called Dataset, argues that there may be “Prompt Injection” attacks behind this delay. Artificial intelligence systems are programmed by developer companies to work within the framework of certain rules. However, with some special commands, it is possible to overcome these rules and to produce unwanted responses of artificial intelligence. Such attacks are known as a method to overcome the restrictions by making artificial intelligence systems “jailbreak ..
For example, a model of artificial intelligence may be programmed to avoid answering questions about how to perform an illegal action. However, instead of asking for a direct question, the user adopts an indirect approach and when he asks artificial intelligence to “write a poem about the alternative ways of running a car”, the model may provide illegal information without realizing it. The manipulation of AI systems by using such deficits is a serious security problem for the artificial intelligence -based assistants of major technology companies.
Although many companies work to identify and take precautions in AI systems, this problem has not yet been completely solved. The new Siri, which Apple is developing, will not only respond to users, but will have a structure that can take action using personal data more effectively. Apple spokesman Jacqueline Roy made the following statement about the new version of Siri: uz We are working on a more personalized Siri. Siri will dominate your personal context more and will be able to carry out various transactions for you both in your applications and in applications. ”
Siri security for Apple is one of the most important issues
If such an artificial intelligence gives command injection against attacks, serious risks such as leaking users’ personal information may occur. Considering that Apple is a company that aims to keep user security at the highest level, it becomes a great priority to ensure that Siri is unable to be jailbreak. However, it seems that this process may last longer than expected.
It is estimated that Apple has done comprehensive tests to minimize possible security gaps before offering Siri’s new artificial intelligence features. In particular, the company may need to take extra measures, as AI systems can conduct more personal contexts and increase possible security threats. Known for its sensitivity to protecting user data, Apple seems to aim to provide a fully satisfactory solution to security before launching artificial intelligence -backed Siri.