Code Ban Twist: Alibaba Flips The Script

Anthropic’s bombshell letter says Alibaba-linked operators hit Claude with 28.8 million covert queries to copy US AI know‑how—and now Alibaba is banning Anthropic’s coding tool across the company as the AI cold war heats up.

Story Snapshot

Anthropic told the US Senate that Alibaba-linked operators ran the largest known AI “distillation” attack against Claude.
The alleged campaign used about 25,000 fake accounts and 28.8 million exchanges targeting advanced coding and reasoning skills.
Alibaba denies using proprietary model outputs for training but is banning Anthropic’s Claude Code, claiming security concerns.
The clash highlights a wider battle over Chinese AI firms siphoning American IP and turning US investment into their shortcut.

Anthropic’s Letter: A Massive AI Capability Grab

Anthropic, the US company behind the Claude AI model, sent a formal letter on June 10, 2026 to the US Senate Banking Committee, chaired by Senator Tim Scott with Senator Elizabeth Warren as ranking member. In that letter, Anthropic claims that operators “affiliated with Alibaba and its Qwen AI lab” ran what it calls the largest known distillation attack against Claude to date. Distillation here means training a rival model on Claude’s outputs instead of building capabilities honestly through years of research. Anthropic warns lawmakers this is not a minor terms-of-service issue, but a direct threat to American technological leadership.

According to the letter, the alleged campaign ran from April 22 to June 5, 2026, a tight 44‑day window timed to harvest maximum value from Anthropic’s frontier systems. During that span, Anthropic says the actors used nearly 25,000 fraudulent accounts—accounts that did not represent real organic users—to generate more than 28.8 million exchanges with Claude. The prompts focused on high‑value behaviors like agentic reasoning, software engineering, and long‑horizon planning, the same capabilities embodied in Anthropic’s advanced Mythos Preview model. In plain terms, Anthropic is telling Congress that a Chinese tech giant tried to copy America’s cutting‑edge AI brain, line by line.

Alibaba’s Denial And Its Claude Code Ban

Alibaba has publicly denied using proprietary US AI model outputs to train its own systems, saying it complies with intellectual property laws and does not rely on rival outputs for training. At the same time, reports say Alibaba is now banning Anthropic’s Claude Code tool across all workplace environments, effective mid‑July, on the claim that the product includes hidden backdoors and fingerprinting mechanisms. So far, Alibaba has not released a technical audit, code samples, or forensic report proving Claude Code secretly tracks Chinese users. That leaves its security claim as an assertion without published evidence.

Anthropic’s letter, by contrast, lays out a concrete pattern of behavior. It describes commercial proxy services used to route traffic and hide the true geographic origin of the accounts, helping operators bypass restrictions that block access from Chinese entities. The usage pattern—huge query volumes from many fake accounts, with prompts narrowly focused on reasoning and coding—fits known markers of a distillation attack rather than normal developer use. Anthropic also notes that this attack is almost twice the scale of earlier campaigns it traced to other Chinese labs like DeepSeek, Moonshot, and MiniMax, which together generated about 16 million exchanges through 24,000 fraudulent accounts. In other words, this is part of a growing playbook, not a one‑off event.

Why This Fight Matters For American Power And Security

This dispute sits inside a bigger pattern: US firms like Anthropic and OpenAI have repeatedly accused Chinese labs of unauthorized distillation, using millions of API calls and thousands of fake accounts to siphon capabilities from American models. Legal experts say this kind of adversarial distillation likely violates terms of service and may be treated as a form of trespass or trade secret misuse, even though existing copyright law struggles to cover it cleanly. For US conservatives, the stakes are straight‑forward. Hundreds of billions of dollars in American AI investment risk becoming a subsidy for foreign rivals who ignore rules and piggyback on US innovation.

Anthropic caught Alibaba using 25,000 fake accounts to steal Claude’s capabilities.

Generated 28.8 million illegal queries in two months.

Told US senators about it.

Alibaba’s response was not a denial.

It was a ban.

Starting July 10, no Claude Code at Alibaba.

No Claude… https://t.co/S81An9Brg8 pic.twitter.com/MYlLACPe8C

— Abdulkadir | Cybersecurity (@cyber_razz) July 4, 2026

The Trump administration has already framed these attacks as both an economic and national security threat, and Anthropic’s letter plays directly into that concern by urging tougher export controls and better intelligence sharing to stop foreign AI IP theft. The US Department of Defense has designated Alibaba as part of China’s military‑civil fusion system, linking the company to broader military goals, even as Alibaba challenges that label in federal court. Lawmakers are now weighing sanctions and new rules targeting foreign firms that “illicitly extract” American AI capabilities. For readers who care about a strong America, secure borders, and fair competition, this AI distillation war is simply the latest front in a long fight to keep US technology out of hostile hands.