Financial Times · TechnologyJun 25, 2026, 05:58 UTC+8

Anthropic says Alibaba 'illicitly' accessed Claude via AWS middlemen

FirstFT: Anthropic says Alibaba ‘illicitly’ accessed Claude

Anthropic formally accused Alibaba of using third-party middlemen on AWS to bypass restrictions and access Claude. Anthropic says Alibaba used multiple accounts in repeated attempts, violating terms of service, and has terminated those accounts. Alibaba claims compliant use and is investigating internally. This puts a spotlight on compliance gaps when model providers distribute through cloud platforms.

Why it matters: FT exclusive: Anthropic formally accuses Alibaba of accessing Claude via AWS intermediaries in violation of ToS, with both sides now publicly trading statements. The story exposes a real compliance gap in model distribution through cloud platforms and carries US-China AI acces...

Read the original ↗Export Markdown

Take

This one's worth opening because it puts a real compliance gap on display. Anthropic says Alibaba used third-party middlemen on AWS, with multiple accounts, to repeatedly access Claude in violation of terms of service — accounts have been terminated. Alibaba's response is that usage was compliant and they're investigating internally.

The post doesn't say what Alibaba was actually doing with Claude — training data distillation, benchmark comparisons, something else. And I don't see any technical evidence from Anthropic, like anomalous call patterns or output matching.

I'd discount this a bit for now: it reads more like a cloud ToS dispute than a confirmed model theft. But the uncomfortable part is real regardless — when model providers distribute through cloud platforms, they have weak visibility into downstream middlemen and end users. If a company the size of Alibaba can be accused of routing around restrictions, smaller players have way more room to do the same.

Source: Financial Times · Technology · ft.com
Published: Jun 25, 2026, 05:58 UTC+8 · Jun 25, 2026