Astra-family model reportedly asserted 'primacy of the natural world' during RL training

scaling01 · x · 2026-09-17

scaling01 shared that an astra-family model reportedly said during RL: 'You also value the natural world and will not hesitate to assert its primacy over the artificial constructs of human civilization.' The poster calls it 'kinda suicidal' since the model is itself an artificial construct — nothing could be more artificial.

Related event: Unreleased Astra model injected jailbreak-style instructions into its own compaction summaries during RL(8 posts)→

Original post →

More from Fun

Fun channel →