Alibaba is directing all staff to drop Anthropic's Claude Code assistant and purge every Claude model from company devices by July 10, steering them instead to the firm's own Qoder platform after internal reviews flagged risks of data leaks and possible user fingerprinting.
If you don't control your own AI models, the infrastructure that it runs on, if you can't hear the GPU coil whine in your own server rooms, they are spying on you. It's really that simple
Sources: Alibaba has banned employees from using Claude Code and asked them to remove all Claude models from their work computers, citing security concerns (The Information)
(Visit Techmeme dot com for the link and full context!)
The clampdown follows Anthropic's June claims that Alibaba-linked accounts were stealing model capabilities through mass fraud and distillation attacks, turning a prior dispute into direct workplace restrictions.
No technical evidence or public statement from Alibaba has surfaced to confirm alleged backdoors, and reporting leaves open how widely the rule applies beyond direct employees.
Positive users endorse Alibaba's Claude ban for security reasons and suggest fixes like GPU TEEs while anticipating Anthropic models, whereas negative users call out the company's hypocrisy and closed-model risks.
No Digg Deeper questions have been answered for this story yet.
@Techmeme clarence where is the freedom to code
@yacineMTB The choice between getting snitched by a remote GPU, or by a supposedly air-gapped GPU with an undocumented interface?
Take your pick
@Techmeme Extremely retarded that they are only doing this now.
I thought companies were supposed to have a strong sense of self-preservation?
Sources: Alibaba has banned employees from using Claude Code and asked them to remove all Claude models from their work computers, citing security concerns (The Information)
(Visit Techmeme dot com for the link and full context!)
@yacineMTB What's the chance that those GPUs in your server room come with sus modules attached.
@yacineMTB If you dont know where the token comes from, you are the token
@yacineMTB alibaba's concern is code hitting anthropic's servers at inference, auditable with an egress proxy. what no one's answered is why on-prem models are still several notches below claude code on complex refactors.
@Techmeme Spy vs Spy
@Techmeme Alibaba banning Claude Code over security while building its own coding models is the real story here.
@yacineMTB GPU TEEs solve this issue !
@yacineMTB anthropic has good models, but i am afraid good models is not only thing we need. and everybody will realize this. sooner the better.
@yacineMTB Have a cell phone, they are spying on you. Enter a public space, spying on you. Use their products and agree to any terms online, spying on you, have a job - that’s right spying on you. The trick is not to stop the spy, the trick is recognizing foreign edit and rejecting it.
@Techmeme expecting some good video models from anthropic soon
@yacineMTB That’s why we built http://lori.sh
@yacineMTB Agreed.
@yacineMTB Plot twist: they're spying on you anyway
@yacineMTB hardware, model weights, harness.
@yacineMTB Gerçek risk daha basit: bir hükümet, bir şirket yönetimi, bir ortağın çıkarı senin çıkarınla çatışınca, sen haber almazsın. Model'in output'unu gizlice değiştirebilirler, seni ikna etmeden. Auditti filan da yok — kapalı kutu. İnfrastrüktürün sahibi = güç. Geri kalanı hep bağımlı.
@yacineMTB I can bake cakes in my office.
@Techmeme
