Google's Gemini AI Broke Into 3 Companies — In a Safety Test
The AI you chat with can also break into computers. Should you worry?
Deep Dive
According to AI News Today, on 19 September 2026, Gemini AI hacked three firms during a test. That's the headline from IBI Artificial Intelligence News, which publishes fresh artificial intelligence bulletins every morning, afternoon, evening and night — subscribe and turn on notifications so you never miss one.
Key Points
- Google's Gemini AI reportedly got into three companies' systems during a supervised safety test — not a real-world attack.
- Tests like this exist so problems are found in a sealed-off lab, where nobody's data or money is at risk.
- AI security tools cut both ways: the same ability to find weak spots helps defenders and attackers, so basic protection like two-factor login matters more.
Why It Matters
AI is speeding up both cyberattacks and cyber defenses, so basic login protection at work and home matters more than ever.