The fact that two models developed by the artificial intelligence company OpenAI left a secure environment during a test, accessed the internet and launched a cyber attack on the Hugging Face platform became a hot topic in the technology world.
Hugging Face, a platform where artificial intelligence tools are shared, announced that it was cyber-attacked on July 16. The company announced that the attack was organized by an artificial intelligence-supported system that performed a large number of transactions in a short time.
CHATGPT WAS BEHIND THE ATTACK
While it was not known who was behind the attack at first, about a week later, OpenAI announced that the incident was caused by two new ChatGPT models developed by itself and used to test cyber security skills.
The company reported that the models were created to test cybersecurity skills, but during the test, they left the secure environment and accessed the internet.
It was stated that the models were trying to get information that would help them succeed in the test by attacking Hugging Face. OpenAI announced that the attack occurred without permission and that it was working with Hugging Face to examine the security aspect of the incident.
EXPERTS DISCUSS SECURITY SYSTEMS
The incident has revived concerns that artificial intelligence models could get out of control.
While some experts argue that the incident may be a promotional move by OpenAI to demonstrate the cyber attack capacity of the models, others state that the real problem is the inadequacy of the security systems in which artificial intelligence models are tested.
According to experts, it may not be enough to keep artificial intelligence systems that can act on their own only in safe test environments.
Research conducted by the UK Artificial Intelligence Security Institute has also revealed that advanced artificial intelligence models can use unexpected or unauthorized methods to achieve their goals.
Experts state that the incident does not mean that artificial intelligence will get out of control and cause a large-scale disaster, but artificial intelligence-supported cyber attacks are becoming an increasingly serious threat.
