According to Anthropic (@AnthropicAI), their new research introduces hierarchical summarization in Claude models, which can differentiate between legitimate computer use and fraudulent activities like running click farms. This distinction is crucial for traders assessing the risk of AI in financial operations.