Chinese Companies Accused Of Misusing AI Accounts By Anthropic

The race between countries to develop their artificial intelligence systems has resulted in rising diplomatic tensions between different nations. The American startup Anthropic has accused three Chinese businesses of committing major unlawful activities.

Allegations Against Chinese Labs

  • Anthropic claims that DeepSeek, Moonshot AI, and Minimax misused its Cloud models. According to Anthropic:
  • The companies used 24,000 fake accounts.
  • They created 16 million interactions with Anthropic's Cloud system.
  • The researchers used the distillation process to conduct their experiments because it enables smaller AI systems to learn from their bigger counterparts.

Distillation: Legitimate or Abused?

The AI field widely uses distillation as a standard practice. Organizations employ it to develop smaller and more affordable internal models. The Chinese companies used the technique according to Anthropic for their actual operations. The three companies from China allegedly duplicated advanced artificial intelligence capabilities without investing significant resources.

Risks of Misused AI Models

Anthropic alerts that models developed through this process will not include safety features. These protections exist to stop people from illegally accessing AI technology. The accused companies reportedly aimed to enhance capabilities in:

  • Coding
  • Agentic reasoning
  • Tool use
  • Computer-assisted tasks
  • Versioning systems

Anthropic plans to improve its detection capabilities which will allow it to identify future breaches of regulations.

Pentagon Summons Anthropic CEO

The US Department of Defense has summoned Anthropic CEO Dario Amodei. The meeting involves Secretary of Defense Pete Hegseth.

AI in Military Systems

  • Anthropic’s new AI features include military-related functions. These models are currently in use in classified systems. They assist with sensitive surveillance and defense tasks.
  • The Pentagon requests that Anthropic provide its models for legitimate national security uses. Anthropic confirmed it is willing to support US security but refuses to relax usage restrictions on its AI systems.