Hacker demonstrates security flaws in GPT-4 just one day after launch

OpenAi's new service, called GPT-4, was recently released to the public. One day after the release of the program, a student was able to discover serious vulnerabilities in the security of the service, possibly exploitable for malicious purposes.

Alex Albert, a computer science student at the University of Washington, posted on twitter a demonstration of how simple it is to ask the program to write the code necessary to hack a computer, thus emphasizing the speed with which security mechanisms of GPT-4 can be bypassed.

The article contains an interview with the student in question, whose discourse focuses on how to promote a broad discussion on the promises and dangers of advanced Artificial Intelligences.