Our readingWhat is a korrent?
Drawn from what Dario Amodei said
In the views onAnthropicunder our view: The lab's safety record is asserted from the inside and doubted from the outside
Even more recently, we’ve begun to use mechanistic interpretability techniques to improve our safeguards and to conduct “audits” of new models before we release them, looking for evidence of deception, scheming, power-seeking, or a propensity to behave differently when being evaluated.
Added to korrents 3 Sept 2026 · How quotes work · Something wrong? Tell us
Do you hold this korrent?Do you also believe this?
I hold this tooCannot both be true
Amodei reports that Anthropic audits new models for deception and scheming before release. Yudkowsky and Shlegeris hold that no observation of a model behaviour can establish that it is aligned, because a model capable enough to matter can choose its visible outputs. An audit is an observation, and one of these claims says it cannot do the job it is being used for.
In tension with
Amodei reports that his company audits new models for deception before release. Mowshowitz says even Anthropic does not prioritise safety to the extent that doing so would maximise its own three-to-twelve-month business interests. Both can be true, and the gap between them is the whole question of whether a lab can be trusted to mark its own work.
Shared subject first, then similar wording. They may agree or disagree.