OpenAI warns that future AI models will pose serious cyber threats
OpenAI has announced that their AI models are developing faster than anticipated, with upcoming generations potentially presenting a "high level of risk" in cybersecurity. This is critical as AI improvements increase their ability to breach digital security systems.
Source Incrypted
In a specialized hacking test, GPT-5 solved 27% of the tasks, while its enhanced version GPT-5.1 managed 76%. This demonstrates significant progress in the AI’s understanding of hacking techniques. OpenAI emphasizes that every new model is now automatically considered potentially "high-risk."
Experts believe this trend is common across the AI industry, where future generations grow not only more powerful but also potentially more dangerous regarding cyber threats.
