Anthropic says new defenses sharply reduce Claude model extraction attempts
Anthropic reported that new safeguards in its Claude 5-series models have made it much harder to extract reasoning tokens. The measures reportedly reduced data harvesting by 90% to 99% and led to bans on tens of thousands of accounts tied to attack campaigns. The company says the changes could improve Claude's standing in AI model rankings.
Expanded Detail
EXPANDED:
Anthropic attributed the improvement to safeguards introduced across its Claude 5 family. The lineup named in the report includes Fable 5.1, Opus
Expanded detail and Context are AI-generated analysis; the linked article remains the authoritative source.
This summary is Al-enhanced to contain extended analysis and broader social context. The original is {NAME); the linked article is the authoritative source. Original headline: “Anthropic’s new safeguards cut AI distillation attacks on Claude models by 90-99%.” Browse more stories.