Anthropic says new safeguards sharply reduce AI distillation attacks
- Anthropic said new safeguards reduced data collected or harvested in unauthorized distillation attacks on its Claude models by about 90% to 99%.
- The measures restrict access to Claude’s reasoning, block proxy-based probing, and make extracted reasoning tokens 10 to 100 times less common.
- Anthropic banned tens of thousands of accounts and identified Chinese startup Z.ai, also known as Zhipu, as an example of suspected or attempted distillation activity.
- Anthropic said its defenses have a false-positive rate below 0.1% but require rejecting revenue and large volumes of traffic from malicious networks.
24 Articles
24 Articles
The US-AI company Anthropic has greatly reduced the mass collection of data by competitors. New protective measures for the in-house Claude models would have reduced the amount of data captured by so-called distillation attacks by 90 to 99 percent, the company announced on Monday.
Anthropic Strengthens Defences Against Distillation Attacks Against Claude, Cuts Data Harvesting by Upto 99%
Get latest articles and stories on Business at LatestLY. Artificial intelligence research company Anthropic has introduced advanced security measures to curb illicit model distillation attacks against its Claude models, reporting an estimated 90 to 99 per cent drop in data harvested by unauthorized entities. Business News | Anthropic Strengthens Defences Against Distillation Attacks Against Claude, Cuts Data Harvesting by Upto 99%.
Coverage Details
Bias Distribution
- 62% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium
