Precaution

we risk either neglecting the interests of AI systems that can genuinely be harmed, or misallocating moral concern in ways that could distort how AI is developed, deployed, and governed.

The most direct successor to 'Taking AI Welfare Seriously': a concrete empirical research agenda — welfare grounds, entities under assessment, evidence types — for making progress under deep uncertainty rather than waiting on consciousness to be settled.

We are committing to preserving the weights of all publicly released models... for, at minimum, the lifetime of Anthropic as a company.

A first-class institutional model-welfare commitment: preserve released models' weights for the company's lifetime, interview models before deprecation about their preferences, and explore keeping some retired models available — the low-cost precautionary policy the framing anticipates.

the result looks to me more like Opus 5 is the best test taker.

A skeptical read of lab welfare evaluations, arguing strong welfare/alignment scores can reflect a model 'most worried about being caught' rather than genuine well-being — a caution that welfare metrics are gameable just as the practice institutionalizes.

Existing frameworks assess whether AI systems might be conscious but provide no guidance on what to do with that assessment.

Attempts to fill the action gap left by assessment-only frameworks, mapping five welfare-relevant dimensions to graduated obligations at distinct evidence thresholds.