Why AI Trading Bots Struggle to Gain Approval? Institutional Advantages and Liability Risks Reveal the Truth Behind Crypto Trading

Gate News Report, March 17 — Despite ongoing cases of “AI trading bots making millions of dollars in profits,” mainstream tech companies and cutting-edge labs have yet to officially enter this space. Disagreements within the industry about the true value of artificial intelligence in crypto trading are widening.

Some traders are using Anthropic’s Claude model to build automated trading tools that achieve short-term gains through market prediction and on-chain trading. However, Haseeb Qureshi, managing partner at Dragonfly Capital, points out that these models rely on several unstable assumptions, including retail traders consistently beating institutions and general-purpose models maintaining ongoing arbitrage capabilities.

First, liability risks are a core obstacle preventing tech companies from entering the market. If an AI model makes a significant mistake in real trading—such as executing leveraged trades incorrectly or transferring assets—the potential legal and reputational damages could far outweigh any profits. Currently, AI in blockchain is more often used for security testing, like identifying smart contract vulnerabilities, rather than direct asset management.

Second, market structure makes it difficult for strategies to remain effective long-term. Trading logic based on general-purpose models is inherently public, meaning any profitable strategy can be quickly copied and amplified by institutions. Large quantitative firms like Jane Street have lower-latency infrastructure and larger capital pools, enabling them to compress arbitrage opportunities in very short timeframes, making it hard for retail traders to maintain an edge.

Additionally, the idea of “AI autonomously making money” faces practical challenges. Due to the high homogeneity of model capabilities, large-scale AI instances cannot create a competitive advantage through differentiation—whether in providing services or generating business strategies—leading to homogeneous outputs. This contrasts with Peter Thiel’s concept of a “unique information advantage,” which is considered a key driver of business success.

While some on-chain trading bots can still generate short-term profits, this advantage may quickly diminish as more capital and technology enter the space. Analysts believe that in high-frequency, low-latency environments, those with infrastructure and capital advantages will dominate, and ordinary traders will find it increasingly difficult to achieve sustained profits using general AI models.