AI Signal 131
Anthropic publishes pilot findings on Claude usage; users delegate high-stakes tasks
↗
Anthropic released findings from a pilot that gave three external researchers access to aggregate Claude usage data, with one study finding that users delegate high-stakes tasks to the model.
This pilot marks a step toward external scrutiny of real-world AI usage, which could inform how AI assistants are evaluated. The finding that users delegate high-stakes tasks suggests Claude is trusted with consequential decisions, raising questions about reliability and oversight. Engineers building on Claude may need to consider safeguards for such delegation.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Anthropic ran a pilot giving three external researchers access to aggregate, real-world Claude usage data.
One study from the pilot found that users delegate high-stakes tasks to Claude.
Anthropic released the findings from this pilot.
THE CLUSTER