Anthropic says its defences cut reasoning-token theft by 99%
What happened
On 13 October 2026, Anthropic said its newly deployed defences cut the volume of reasoning tokens extracted from Claude by 90% to 99%. According to the company, such attacks use networks of hundreds of thousands of fake accounts to systematically copy model capabilities in order to train smaller models of their own. Anthropic said the new defences target precisely this distillation-style theft, and gave the reduction range above. Reports so far mention only that the defences are live and the range of their effect; no technical details, number of affected accounts or identities of the attackers were disclosed.
Written by AI from the reports · updated 3 d ago
Timeline
Follow the reports to see the event from each side.
- The DecoderAnthropic says it cut Claude reasoning token theft by up to 99% in fight against industrial-scale distillation
Anthropic says it has cut the extraction of Claude's internal reasoning tokens by 90 to 99 percent. The company links the large-scale distillation attacks to Chinese companies Zhipu and Moonshot, and has already banned tens of thousands of accounts.
Heat of this event
Current heat 3·Peak in the comparable range 9（Oct 13 23:00）·Change over 24 hours in the comparable range -50%
The trend compares only the same accounts observed without a break, so its range can be narrower than the current heat count. Move the pointer or tap the chart to see each hour; with a keyboard, use the left and right arrow keys.