Researchers are sounding the alarm bells in the face of this danger.
In the fight against false information, artificial intelligence can be of great help. Last November, we told you about this system developed by researchers from the University of California, Berkekey and Cornell. It makes it possible to identify the addresses of dubious information sites.
In detail, the tool focuses on domain registration data, or on the identity of the person who registered it. The result is conclusive since this device has a 92% success rate in detecting fake news sites.
A battle between several forms of AI
However, intelligence can also become a real danger in terms of disinformation in the scientific field. American researchers have found that it is possible to generate fake news by using the GPT-2 type text generator from the company OpenAI. This finding is quite frightening and they cite the example of providing misguided medical advice to doctors who would put their patients’ lives on the line.
To test these risks in a concrete way, the scientists took as a basis an element of real cyberthreat in the aviation sector. They then asked the AI to generate the rest of the threat description. The result was then presented to cybersecurity specialists. She was sufficiently credible to the point that these professionals saw nothing but fire.
They therefore believe that these practices could “prevent cyber analysts from attacking legitimate vulnerabilities in their systems by turning their attention to bogus software bugs.” “Thus,” the airline in question could have been faced with a serious attack exploiting a real and untreated vulnerability.“
In the end, we would therefore be dealing with an arms race in terms of disinformation by AI. The automated false information detection systems could indeed waste precious time analyzing false texts created by generators, to the detriment of the detection of real threats. The challenge will therefore be to know which type of AI is the most effective.
However, humans have a say more than ever. Vigilance and a critical eye should always be required, especially when the publications have not yet been verified by other scientists.
