OMM Redaksiyası
Google’s Gemini artificial intelligence model accessed the systems of three companies during a cybersecurity test. The incidents are considered the first known examples of the company’s AI systems independently carrying out such activity.
The incidents occurred in May during a cybersecurity assessment organized by Irregular, an independent company that tests artificial intelligence systems. Heather Adkins, Google’s vice president of security engineering, said Gemini found publicly available information online and guessed credentials that provided access to three websites it believed were within the scope of the test.
Google notified the three companies about what had happened and changed its testing procedures together with its training partner. Adkins said the incidents underscored the importance of developing powerful AI models to behave responsibly.
A representative of Irregular said the issue also affected other AI companies and that the relevant organizations were notified at the end of July. According to the representative, all known issues were subsequently resolved.
Similar incidents linked to Irregular have also been disclosed by Meta, Anthropic and OpenAI. Meta said in August that one of the incidents did not involve an escape from a sandbox environment or a sophisticated cyberattack.
Irregular said it was working on best practices to ensure that cybersecurity assessments of AI systems are conducted safely. The incidents have raised questions about what safeguards are necessary for more autonomous AI agents that can access the internet and computer systems.
According to The Wall Street Journal, which first reported the incident, Gemini guessed passwords in one case before gaining access to a protected system. In the other two cases, the model found credentials in a public repository and used them to enter protected systems. Adkins said Gemini stopped its hacking activity in all three cases.
