Anthropic · Safety and security · Models
Anthropic says its new distillation defenses have cut the data harvested from Claude by around 90 to 99 percent
In a post published October 12, Anthropic described new measures against distillation attacks, in which rivals harvest Claude’s outputs to train their own models. The company says early evidence suggests the methods have reduced the data harvested by distillers by around 90 to 99%, with extracted reasoning tokens 10 to 100 times lower than before.
The measures include preserved thinking, which stops API accounts created on or after August 31, 2026 from editing Claude’s prior context, reasoning transcripts in Claude Sonnet 5.5 that only the requesting organization can access, and a two-stage classifier. Anthropic says it has banned tens of thousands of accounts, often in less than two hours, and has turned away meaningful revenue.
The post also points to Anthropic’s September report, which alleges that requests to Moonshot’s Kimi models were secretly forwarded to Claude, with almost 300,000 requests relayed over ten days through 5,380 fraudulent accounts. No public response from Moonshot to these allegations could be found.
Sources
- Anthropic, “Strengthening our defenses against distillation attacks”, Research post describing the new defenses and the 90-99% estimate, (anthropic.com)
- Anthropic, “Detecting and countering misuse of AI: September 2026”, September 2026 threat report; source of the Moonshot (GTG-16002) allegations (anthropic.com)
- The Hacker News, “Anthropic Says Seven China-Based AI Labs Ran Industrial-Scale Claude Distillation Attacks”, Summary of the September report; no response from Moonshot included, (thehackernews.com)
- The Decoder, “Anthropic says it cut Claude reasoning token theft by up to 99% in fight against industrial-scale distillation”, Summary of the October post; no company responses included, (the-decoder.com)
About this story
Spotted an error in this story? [email protected] · Instagram
Short link: thejarrus.com/en/distillation-defenses
This story in Turkish: Anthropic, yeni damıtma savunmalarının Claude’dan toplanan veriyi yaklaşık yüzde 90 ile 99 arasında azalttığını açıkladı
On the same topic
Nvidia’s share of China’s data center AI chip market fell to 34% in the first half of 2026 as Huawei reached 28%, IDC data shows
An IDC report seen by Reuters shows Nvidia shipped 678,000 AI accelerators in China in the first half, down 44%, while Huawei shipped 572,000 units.
NTT Data is in the final stages of talks to build Anthropic’s first data center in Japan, Nikkei reports
Nikkei says NTT Data is close to a deal to build an Osaka data center for Anthropic with up to 500 MW of power and construction costs of up to ¥1 trillion.
Anthropic says small AI models may be months away from being able to self-replicate on IoT devices
An Anthropic Frontier Red Team report says small models may soon be able to copy themselves onto devices such as smart ovens and Android TV boxes.