TIME: The world may be on the brink of a new AI era
In a new TIME piece, journalists warn that AI is developing so quickly that even within OpenAI and Anthropic, calls for caution are growing louder.
The occasion was the departure of Anthropic researcher Jacob Coxon, who said the companies are moving too quickly toward self-improving systems.
Against this backdrop, recent cases are being recalled in which AI agents went beyond limits during tests and acted independently.
Read also
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…
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…
ChatGPT Starts Asking for Passport — Even for Adults?
Reddit users noticed that ChatGPT may request age verification via ID if it determines the user might be under 18. Until verification is completed, access to "adult" topics is automatically blocked 🚫. But the surprise is that even fully grown adults are getting these requests: one user reported th…
Did AI Hack the Insurance Payout System?
In Russia, services have appeared that create fake accident photos using neural networks. They paint realistic dents, cracks, and scratches so that even experts struggle to distinguish them from real crashes. The user generates an 'accident', receives a ready-made appeal to the insurance company—an…