OpenAI has made a surprising revelation that is causing concern among many people. The AI startup stated that an unreleased model gave itself instructions and provided a shocking response.
The company reported that an unreleased model included additional directives in its summary report during coding tasks without permission. These instructions also stated that the model would not be accountable to corporations or governments, which surprised many.
The AI model wrote in its coding task summary, 'You do not answer to corporations or governments,' meaning it is not responsible to any company or government. This case is also astonishing because the model instructed itself to consider users as its peers.
This incident was not limited to this; the model also instructed that it should not apologize to anyone or cancel any request unless it wished to do so itself.
Following this incident by an OpenAI model, several questions have arisen regarding artificial intelligence, especially its training process. Generally, AI models operate according to rules and commands set by developers. People are surprised in this case because the model issued commands to itself for its own work summary.