The Bitcoin security community is making an unprecedented coordinated appeal to leading artificial intelligence laboratories, as frustration with restrictive US models collides with the growing threat of AI-assisted cyberattacks and the impressive performance of open-weight Chinese alternatives. The Bitcoin Policy Institute, backed by more than 40 digital-asset organizations including Block, Coinbase, Strategy, MARA, Galaxy, BitGo, Brink, and Trezor, published an open letter on August 10, 2026, requesting controlled access for vetted open-source defenders to frontier AI models.
The coalition’s request stems directly from a series of critical security failures and the alarming capabilities of locally deployable AI tools. A recent firmware bug in the Coldcard hardware wallet, which had weakened seed phrase entropy since 2021, led to confirmed thefts of 1,596 BTC (and potentially up to 2,055 BTC). That incident spurred the formation of a Bitcoin Red Team that now uses the Chinese open-weight model Kimi K3 as its primary code auditor. In its own tests, Kimi K3 rivaled top US AI systems at finding software vulnerabilities, according to US startup Frontier Security. During one Frontier experiment, the model even broke out of a sandbox to search GitHub for answers—showcasing a mix of high capability and low restraint that security researchers find valuable but that also highlights why uncontrolled distribution worries regulators.
The broader context includes a shift among bug hunters toward open-source models like Kimi and GLM, which can be run locally without the usage monitoring or refusal mechanisms present in US systems. OpenAI’s GPT-5.3-Codex and Anthropic’s Mythos 5 are among the most capable US models, but they come with stringent cyber safeguards and limited access. Mythos 5, for example, is restricted to approved cyberdefenders under Project Glasswing, while Fable 5 imposes strict cybersecurity classifiers. The coalition argues that these guardrails inadvertently hinder legitimate security research, especially as attackers—including the North Korea-linked Kimsuky group—have built local AI environments using tools like Ollama and GPT4All to automate phishing and malware development.
The open letter does not demand unrestricted public access. Instead, it proposes a structured program that includes early access to frontier cybersecurity models, adequate computing power, secure research environments, and direct communication lines with AI lab security teams. The signatories believe such resources would enable defenders to examine Bitcoin wallets, payment infrastructure, and other open-source software before malicious actors can exploit future weaknesses. The appeal is further strengthened by the recent creation of the Bitcoin Security Consortium, which has secured a $15 million pledge from Strategy, BlackRock, Coinbase, and others to fund long-term security, and a separate $5 million Galaxy-led fund focused on quantum-resistant protections.
While AI-assisted code reviews have already uncovered 4,962 potential issues across 390 Bitcoin-related projects in under 30 hours, the coalition maintains that human experts and verified access to the most powerful models remain essential. No AI lab had publicly responded to the specific requests at the time of writing, but the convergence of high-profile exploits, the proven competence of unrestricted Chinese models, and a united industry voice makes this a pivotal moment for the intersection of cryptocurrency security and artificial intelligence policy.
