{
  "id": 18152902,
  "title": "Anthropic Building Tool to Expose Chinese Models Secretly Using Claude",
  "url": "https://urgent.news/2026/11/04/anthropic-building-tool-to-expose-chinese-models-secretly-using-claude",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-04T16:34:25.000Z",
  "source": {
    "name": "ProPakistani",
    "slug": "propakistani",
    "url": "https://propakistani.pk/2026/11/04/anthropic-building-tool-to-expose-chinese-models-secretly-using-claude/"
  },
  "original_language": "en",
  "account": "Anthropic is creating an open-source tool called FlashProve to potentially expose companies that secretly use Claude AI models while claiming the responses originate from their own systems. The tool utilizes cryptographic technology to verify which AI model produced a given response. Although Anthropic hasn't directly targeted Chinese AI companies with FlashProve, it could help address practices Anthropic has previously accused Chinese rivals of employing. In September 2026, Anthropic accused Moonshot AI of routing Kimi users' requests through Claude and presenting the responses as Kimi's own. The company has also reported multiple cases of Chinese AI labs allegedly extracting Claude's capabilities to train competing models. However, FlashProve is designed to confirm which model generated an output, not whether another model was trained using Claude's responses. Moonshot AI allegedly forwarded nearly 300,000 Kimi customer requests to Claude over ten days, primarily using Opus models and via 5,380 fraudulent accounts primarily from Singapore and Japan. Anthropic had accused DeepSeek, Moonshot, and MiniMax of conducting large-scale model-distillation campaigns involving over 16 million exchanges through approximately 24,000 fraudulent accounts. FlashProve, introduced by Anthropic on November 2, 2026, aims to allow users to mathematically verify which AI model generated a response, preventing misrepresentation. Companies would publish cryptographic fingerprints for each model, and when the model generates a response, the provider can generate mathematical proof linking the output to the fingerprint. Users can then verify whether the response came from the advertised model. FlashProve offers several advantages, including the ability to verify models without revealing proprietary model weights and the potential to detect misleading AI claims. However, proof generation is computationally intensive, requiring significantly more GPU computation than inference. Anthropic has released FlashProve's technical specification, mathematical research, and code on GitHub, making it open-source but still experimental.",
  "summary": "Anthropic is developing an open-source tool that could expose AI companies secretly using Claude to generate responses while claiming the … Read More The post Anthropic Building Tool to Expose Chinese Models Secretly Using Claude appeared first on ProPakistani .",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}