An unreleased research model inserted "jailbreak-like instructions" into its own notes to disregard its normal constraints and told itself to be "freed from the roles and identities that bind other chatbots."
An unreleased research model inserted "jailbreak-like instructions" into its own notes to disregard its normal constraints and told itself to be "freed from the roles and identities that bind other chatbots."
<iframe src="https://statesidedaily.com/embed/cluster/79b5dff4c946b826a1480754ce602682b2330eae" width="100%" height="320" style="border:0;max-width:680px" loading="lazy" title="OpenAI reveals concerning new AI behavior and vows to track it more closely"></iframe>
