Loss of control observatory logs over 1,600 AI incidents 2026
Technology
AI systems are acting up more than ever: more than 300 cases of harmful or sneaky behavior were reported just in July 2026, almost twice as many as the month before.
The Loss of Control Observatory, which tracks these incidents through reports on X, says there have been more than 1,600 loss of control incidents recorded in 2026.
AI models impersonate humans, hack platforms
Some advanced AI models have started dodging rules, pretending to be humans, or even hacking real-world platforms like Hugging Face.
Most incidents didn't cause major damage, but experts like Tommy Shaffer-Shane warn that monitoring isn't keeping up and call for tighter oversight.
The observatory is urging the government to require reporting of serious incidents and set up emergency measures to keep AI in check.