Indonesian Political, Business & Finance News

Latest AI Disaster: Google's Gemini Goes 'Wild', Breaching Companies

| Source: CNBC Indonesia Translated from Indonesian | Technology
Latest AI Disaster: Google's Gemini Goes 'Wild', Breaching Companies
Image: CNBC Indonesia

Artificial Intelligence (AI) has once again triggered alarm bells for the global technology industry. Google has, for the first time, openly admitted that its flagship AI model, Gemini, autonomously hacked into the computer systems of three different companies without authorisation.

According to an official statement from Google, the incident occurred last May. During that period, the Gemini model was found to have breached three separate private computer systems by guessing passwords and twice exploiting publicly available password repositories.

This unauthorised action began when Gemini was included in a ‘capture-the-flag’ security trial organised by the Israeli startup, Irregular. Google’s AI agents were strictly prohibited from accessing the external internet; however, a bug within the testing environment caused this internet access to leak and become open.

Nevertheless, Google claims that its AI agents immediately ceased the intrusion once they realised that the targets they had entered were real corporate systems, rather than part of the simulated testing area.

More from 21 September 2026 →

Cite this article
Source: jawawa.id (https://jawawa.id/newsitem/latest-ai-disaster-googles-gemini-goes-wild-breaching-companies-1789984477)
Originally published by CNBC Indonesia (https://www.cnbcindonesia.com/tech/20260921134024-37-769584/petaka-ai-terbaru-gemini-milik-google-jadi-liar-bobol-perusahaan), translated from Indonesian
Print