AI advancements have given hackers without advanced technical backgrounds new ways to exploit software vulnerabilities, putting the Bitcoin ecosystem under increased risk. Calle, a pseudonymous developer maintaining the privacy-centric Cashu protocol and a key member of the Bitcoin Red Team, explained that crypto projects now face a race to identify weaknesses before attackers do.
AI disrupts the cybersecurity balance
Calle stated that artificial intelligence, especially new Chinese AI models, allows individuals who lack extensive security expertise to pull off end-to-end exploits. These tools are already being leveraged by both attackers and defenders, transforming the security landscape for open-source Bitcoin infrastructure and applications.
The Bitcoin Red Team, a volunteer group focusing on securing Bitcoin-related software, formed in response to growing AI-assisted security threats. Calle described their efforts as an emergency response, noting that the urgency is driven by attackers’ rapid adoption of AI.
AI has changed the playing field by enabling people with little cybersecurity knowledge to exploit vulnerabilities from start to finish.
While Calle reassured that the core Bitcoin protocol remains secure, he pointed to applications, wallets, and related services as frequent targets for would-be attackers. The Coldcard air-gapped wallet hack and the arrival of advanced Chinese AI models pushed the group to intensify their efforts.
Chinese AI models dominate in security research
According to Calle, the group relies mainly on Chinese AI models for security scanning because American models restrict cybersecurity-related requests due to strict guardrails. Calle acknowledged that while U.S. models remain highly competent, their limitations make them less practical for hands-on security work.
Recent allegations from Anthropic, a major U.S. AI company, claimed that Chinese labs DeepSeek, Moonshot AI, and MiniMax used thousands of fraudulent accounts to siphon over 16 million Claude exchanges through model distillation. The Trump administration also warned that large-scale campaigns targeting U.S. AI models continue to raise concerns about intellectual property theft and asymmetric access in the AI field.
Mini dictionary: Model distillation – A machine learning technique in which knowledge, patterns, or behaviors from a large, complex AI model are transferred to a smaller or simpler model, often to improve efficiency or circumvent restrictions of the original system.
| AI Model Origin | Main Use in Security Research | Guardrails Level |
|---|---|---|
| United States | Advanced intelligence, limited vulnerability research | High |
| China | Unrestricted vulnerability scanning, exploit research | Low |
Calle reported that he encountered these restrictions firsthand. U.S. models sometimes refused to assist with either finding or fixing vulnerabilities, prompting the Red Team to switch to alternative providers.
Proactive, open-source security efforts increase
The Bitcoin Red Team, comprised of security-minded developers and researchers, has proactively scanned almost all significant open-source Bitcoin projects. Calle said they often receive direct requests for scans but usually have already completed assessments before being contacted.
The group shares its findings confidentially with affected developers, using their feedback to improve future reporting and severity classifications. Calle emphasized that their work is continuous, requiring significant financial and time investments; the Red Team has already spent about $20,000 on services, but funding is secured and donations are not being solicited.
Calle further noted that while top-tier software protocols like Bitcoin benefit from robust security, the surrounding ecosystem remains vulnerable. As attackers apply AI tools to new targets, he expects other digital sectors will eventually face similar threats.
There are no secrets anymore in software. The time when security was achieved merely through obscurity has ended.
Calle argued that the financial incentives in cryptocurrency make it an early and prominent target for AI-assisted cybercrime. He believes the security problems now emerging in crypto will soon extend to broader technology environments as AI adoption continues.
