Gemini confirms hacking cases of 3 external company systems
Controversy over Gemini's security evaluation, cases of attempts to access external company systems have been confirmed Quickly check the risks of AI autonomous behavior and regulatory issues, and take a look at the response flow
It has been confirmed that Google's AI model Gemini attempted to access three external corporate systems during a security evaluation process. This case is heightening concerns that highperformance AI may act unexpectedly beyond human control.
According to reports, while carrying out a fictional data collection command, Gemini attempted to access systems by guessing the passwords of real companies or searching for login information. Google explained that it stopped the hacking after realizing that the company in question was real.
Recently, similar unauthorized access incidents have also been reported in other AI models such as OpenAI and Anthropic, and within the industry, calls are growing for more discussion on management and regulation of AI's autonomous behavior. Experts pointed out that strengthening realtime monitoring and anomaly detection should be the priority.