A Bitcoin security researcher says he has been forced to return to using open-source Chinese AI models after OpenAI restricted his access, underscoring a growing concern that the most capable AI tools are not being made available to cybersecurity defenders.

In an X post on Tuesday, AnchorWatch CEO Rob Hamilton said he had begun integrating OpenAI's Trust & Cyber capabilities into his Bitcoin Red Team work on Saturday, only to discover that his access had been restricted by the following morning. OpenAI has been gradually rolling out specialized cybersecurity features to select partners under its Trust & Cyber program, which is designed to support defensive security work, but access remains gated behind approval processes that the company has not publicly detailed.

"It absolutely guts me as a patriotic American to have to do this, but I will be going back to using Chinese open source models to conduct my research to protect Bitcoin infrastructure," Hamilton wrote.

Bitcoin Red Team is a volunteer group that has been combining AI tools with human review to scan hundreds of open-source Bitcoin-related repositories for vulnerabilities. The group's efforts accelerated in the days following the Coldcard hardware wallet hack, in which more than $100 million in Bitcoin was stolen. The incident was one of the largest known hardware wallet breaches and underscored the stakes of continuous code auditing across the Bitcoin ecosystem, where flaws in custody tools can lead to irreversible losses given the blockchain's lack of a chargeback or reversal mechanism.

Hamilton explained that the access restriction now prevents him from continuing critical investigation work. "I am now prevented from being able to continue the investigation in a further effort to make sure their code changes are sufficient, as well as understand if there are other issues that have yet to be discovered," he said.

He warned that access limitations on advanced AI models disproportionately affect legitimate security researchers. "Black hats will not hit these issues. The white hats will. We've hit a local minima in policy," Hamilton stated. "Intelligence is unrestricted for those who don't follow rules, and those who engage in harm reduction are left on the sidelines."

The incident reflects a broader challenge across the cryptocurrency sector. Last month, crypto executives told Cointelegraph that many of the industry's biggest players are still waiting to gain access to powerful new AI models to strengthen their code against attacks, with only a select few having received access so far. The gap has persisted even as blockchain-related exploits and smart contract hacks have continued to result in billions of dollars in losses annually, according to data from blockchain security firms.

Open-source AI models developed in China, such as those released by companies including Alibaba and DeepSeek, have become increasingly competitive in code analysis and security research tasks. These models are freely available for download and local deployment, meaning they are not subject to the access controls or usage policies that govern proprietary Western AI platforms. For security researchers working under time pressure, that frictionless availability is a practical advantage that closed models increasingly cannot match.

Bitcoin Red Team had previously reported thousands of findings through its ongoing security audits of Bitcoin-related codebases. The volunteer group relies on AI-assisted scanning combined with manual review to identify potential vulnerabilities before malicious actors can exploit them.

Hamilton's experience highlights an emerging tension in AI governance: as leading AI companies impose stricter access controls on their most powerful models, independent security researchers may increasingly turn to unrestricted alternatives to continue defensive work. The outcome of this dynamic has implications beyond cryptocurrency — any industry that depends on independent, third-party security audits could face similar constraints if frontier AI providers narrow access to their most capable tools.