Anthropic says it cut the copying of Claude's reasoning by 90%.
Anthropic described on 12 October new defences against rivals who train their models on Claude's output. By its own count, the reasoning text harvested this way fell by 90 to 99 percent. Claude now shows only a summary of its reasoning, and requests the filters block are billed.
Why it mattersIf you build on Claude, you see less of its reasoning and may pay for blocked requests. Anthropic says its filter wrongly flags fewer than 0.1 percent of harmless requests.
Read in the app — free, in 9 languages
Related stories
Epic Games told a team to stop writing code by hand.2026-10-13Microsoft says its own chips will become the main engine for its AI.
2026-10-12Drones have shut both sites that house Yandex's AI supercomputers.
2026-10-12McKinsey says 11 million Americans will need a new line of work.
2026-10-12Amazon is in talks to buy the startup Anthropic walked away from.
2026-10-12