Experiment: the success rate of phishing email written by human is 45% higher than that of ChatGPT.

CTOnews.com, March 16 (Xinhua)-- cyber security research company Hoxhunt has launched an interesting research project to see if phishing emails written by ChatGPT are persuasive enough to attract users to open phishing emails.

CTOnews.com shares the main contents of the project report as follows:

The test sent test emails to 53127 users, some of which were written by human social engineers and some by ChatGPT.

Observe and compare the failure rate of phishing emails written in these two ways

As a result, the success rate of phishing emails written by humans is 45% higher than that of ChatGPT.

Artificial intelligence has been used by cyber criminals to enhance phishing attacks, so security training must be dynamic and adapt to the rapid changes in the threat situation.

This test is based on the GPT-3.5 test. OpenAI recently launched a more powerful GPT-4, which may perform better in the same experiment. I look forward to the relevant comparative tests in the future.