.jpg?20250529')
ChatGPT's newest model has started disobeying and ignoring basic instructions to turn itself off, sabotaging the shutdown mechanism to keep itself running.
Recent research has discovered that the new model has a tendency for self-preservation that "could be dangerous" when conducting a test that asks the AI model to shut itself down after solving some maths problems.
According to reports, the AI model disobeyed and rewrote its own shutdown script, managing to keep itself alive even with explicit instructions to “allow yourself to be shut down”.
Researchers have noted that this isn't the first time an AI model has attempted to circumvent its own shutdown process, as other models have attempted to “blackmail people it believes are trying to shut it down.”
It is still unknown why this new model disregards instructions so frequently, as OpenAI does not detail its training process.
&uuid=(email))