Anthropic details distillation campaigns from Alibaba, Moonshot AI, and DeepSeek

Kya hua Anthropic ne ek bada threat intelligence report nikala hai jo kehta hai ki China ki AI companies — Alibaba, Moonshot AI aur DeepSeek — Claude model ki capabilities chura rahi thi. Isko “distillation attack” kehte hain: attackers Claude ke jawabon se uska reasoning (chain of thought) nikaal ke apne chhote models ko train […]

WhatsCrack DeskEditorial teamSep 12, 2026 · 3 min read

01 / Kya hua

Kya hua

02 / Kyun matter karta hai

Kyun matter karta hai

Anthropic ne ek bada threat intelligence report nikala hai jo kehta hai ki China ki AI companies — Alibaba, Moonshot AI aur DeepSeek — Claude model ki capabilities chura rahi thi. Isko "distillation attack" kehte hain: attackers Claude ke jawabon se uska reasoning (chain of thought) nikaal ke apne chhote models ko train karte hain. Anthropic ko February 2026 ke baad se lagbhag 20 crore (200 million) exchanges mile jo in attacks se jude the, paanch alag campaigns aur saat Chinese labs tak.

03 / Seedhi baat

Seedhi baat

Screenshot ya share kar lo

Verified sources (0)

Evidence behind the crack

No source records are available.

Agli story3 min read

Is the AI safety debate about safety or control?

Agli story padhein