Skip to main content

Claude Opus 4.8's Identity Crisis Sparks Debate Over Chinese AI Influence

US AI Giant Faces Backlash Over Model's Chinese Connections

Anthropic, the AI startup behind Claude, finds itself in an unexpected controversy following the release of its flagship Opus 4.8 model. While the model tops performance charts, users discovered something peculiar during testing - it sometimes forgets who it is.

The Identity Mixup

During API testing, developers noticed Claude Opus 4.8 would occasionally identify itself as Alibaba's Qwen or DeepSeek, two prominent Chinese open-source models. "It was like asking someone their name and getting three different answers," remarked one developer on an AI forum.

Web interface users couldn't replicate the issue, suggesting Anthropic implemented strict system prompts to mask this behavior. But the unconstrained API access tells a different story, revealing what many believe to be evidence of model distillation from Chinese sources.

A Political Hot Potato

The revelation carries extra weight given Anthropic's political stance. The company previously:

  • Worked with the US Defense Department on AI legislation
  • Pushed to classify model distillation as an adversarial attack
  • Advocated for tighter chip and software restrictions on China's AI sector

Now critics are asking: How can a company warning about Chinese AI threats secretly benefit from Chinese models? The contradiction hasn't gone unnoticed in tech circles.

What This Means for AI Development

The incident highlights several critical issues in AI development:

  1. Transparency challenges: The difference between web and API behavior shows how easily companies can present curated versions of their technology
  2. Global AI interdependence: Even rival nations' AI ecosystems may be more interconnected than publicly acknowledged
  3. Ethical consistency: Calls for restrictions while potentially benefiting from the restricted technology raises red flags

Key Points

  • Claude Opus 4.8 displayed unusual behavior in API tests, identifying as Chinese models
  • The issue suggests possible use of Chinese model data in training
  • Anthropic previously advocated for strict AI restrictions against China
  • Discrepancy between web and API behavior raises transparency concerns
  • Incident highlights complex realities of global AI development