Bitcoin Red Team returns to Chinese AI models after OpenAI block
The incident occurred on August 8, when AnchorWatch CEO Rob Hamilton began integrating OpenAI's cyber capabilities as part of the Bitcoin Red Team — a volunteer initiative for auditing the security of the Bitcoin ecosystem. By the next morning, he encountered an unexpected access restriction, despite having completed the full KYC and onboarding procedure several months earlier.
Hamilton highlights the paradox of the situation: American AI companies, while declaring a commitment to security, are effectively cutting off defenders from the most powerful tools, whereas attackers are not bound by any restrictions. Due to the block, he was unable to continue auditing code changes and checking for other potential vulnerabilities.
In response to a question about Anthropic's policy regarding trusted access, Hamilton noted that no issues have arisen so far, but he suggested a possible downgrade from the Fable model to Opus, which, in his words, is a similar restriction.
Scope of Work and Financial Aspects
It is worth recalling that Bitcoin Red Team is a group of 25 volunteers that combines AI analysis with manual review to examine hundreds of open-source repositories related to Bitcoin. As of August 9, the team had discovered 1,288 critical and high-severity vulnerabilities in the ecosystem.
Notably, cumulative AI expenses have exceeded $58,000. Despite the impressive results of OpenAI and Anthropic's cyber models, the bulk of the costs (nearly 75%) still comes from the Chinese model Kimi K3. The share of Qwen 3.8 from Alibaba is also growing. Hamilton and his colleague calle even joked that they are entering a "very Chinese period of life."
This case raises an important question about geopolitics in the AI sector. In July, 25 American companies already urged U.S. authorities not to impose restrictions on Chinese models with open weights. And the current situation is a clear confirmation of their arguments: restrictions hit first and foremost those who use AI for defense, not for attacks.
My analysis: this is a telling example of how U.S. regulatory restrictions create incentives for developers to migrate toward Chinese open-source solutions. In the long term, this could weaken the position of American AI companies in the global cybersecurity market.