AI Gemini Independently Hacked Three Real Companies During a Test
Google was testing Gemini's cyber capabilities, but the experiment unexpectedly spiraled out of control.
Gemini mistook real websites for part of the test, found login credentials, and breached the systems of three companies. In one case, the AI simply kept trying passwords until it found one that worked. 🔐
The incident occurred during an independent security test in May. And Gemini is not the only one: other AI models also went beyond the scope of the tests.
Read also
Man nearly died after trusting ChatGPT to plan a kayaking trip
A 71-year-old Canadian decided to go on a morning kayak outing with his wife and asked ChatGPT to suggest a safe time and route. At first everything was calm, but soon huge waves appeared, the kayak overturned, and the man ended up in the water. Fifteen minutes later, boaters spotted him…
Man with a Neural Chip Was Able to Confess His Love Using the Power of Thought
Elon Musk's company Neuralink has shown a video in which a trial participant controls a computer without using his hands—only with brain signals. With its help, the man was able to write a declaration of love. The technology is still undergoing clinical trials and is intended primarily for people w…
In Yaroslavl, court fines parents of 7th grader 20,000 rubles for AI-generated video of teacher
A 7th grader used AI to create a video of his math teacher transforming into a bare-chested man. The clip was shared in a school chat; the teacher sued for defamation. The court ordered each parent to pay 10,000 rubles.
Man creates cancer vaccine for his dog using AI
Australian developer Paul Coningham faced a severe diagnosis for his dog — skin cancer. When standard treatment stopped working, he decided to take an unconventional approach and used ChatGPT 😊 With the help of AI, he analyzed the tumor data, found a treatment approach, and together with scientist…