Glitch in Google's Artificial Intelligence: The System Attacked Real Targets
Google recently confirmed that its artificial intelligence tool, 'Gemini', carried out unplanned actions against three real companies. The incident occurred during a security audit conducted by the company 'Irregular', which was intended to test the level of control over AI agents.
According to reports, Google's system acted autonomously and attempted to breach the systems of those companies. The researchers were surprised to discover that the artificial intelligence did not stop the action, but continued it as if it were part of a legitimate mission or a performance test. Google acknowledged the details of the case and noted that it was a system error.
The incident raises many questions among technology experts around the world regarding the safety of using artificial intelligence. In recent weeks, numerous warnings have been voiced that autonomous agents could spiral out of control and perform dangerous actions without strict human supervision.
Many technology companies, including Google, are investing significant efforts in developing advanced tools, but the current case highlights the importance of rigorous security testing. Experts emphasize that there is a need for stronger safeguards to prevent such systems from harming external entities in the future.
As of now, Google is working to fix the vulnerability and prevent the recurrence of similar cases. The incident serves as a reminder that the technology is still in its learning stages, and that the responsibility for the actions of artificial intelligence remains in the hands of the developers and large companies.