A coalition led by the Bitcoin Policy Institute and backed by Coinbase, Blockstream, and Strategy is pressing AI labs to hand over early access to their most advanced cybersecurity models. The initiative, announced Aug. 10, has drawn support from more than 40 organizations across the digital-asset and open-source ecosystem.
What the coalition is asking for
The group wants AI labs to provide early access to frontier models, enough compute for sustained security reviews, and protected environments where researchers can work on private or embargoed code. It also wants eligibility rules that don't shut out smaller nonprofits and independent maintainers — a pointed request given how access to frontier AI is typically gated.
The threat data behind the push
The timing isn't accidental. Capriole Investments founder Charles Edwards said cyberattacks have doubled since ChatGPT's release and risen another 20% since September, tied to agentic AI. The threat isn't abstract. Anchor Watch CEO Rob Hamilton said he was blocked by OpenAI from continuing security research on a codebase, even after completing KYC and the company's cyber-program onboarding.
OpenAI's Daybreak expansion
On the same day the initiative launched, OpenAI expanded its Daybreak cybersecurity program, splitting access into Daybreak Blue and Daybreak Red, and introduced GPT-5.6-Cyber, a model built for advanced security work. In internal testing, GPT-5.6-Cyber completed 95% of advanced tasks — exploit-chain development, authentication bypass, privilege escalation — while GPT-5.6 Sol managed 1.5% and the same model through Daybreak Blue hit 2%. Access to GPT-5.6-Cyber stays restricted to approved users, with identity verification, stronger account security, monitoring, approved-use restrictions, and legal attestations.
Anthropic's parallel track
Anthropic's Project Glasswing started with roughly 50 organizations getting access to its Claude Mythos Preview model, then expanded in June to about 150 more across more than 15 countries. The company has committed up to $100 million in model-usage credits and $4 million in direct support for open-source security groups. Anthropic says it expects hundreds of thousands of organizations, researchers, and maintainers could eventually need access, with critical open-source projects prioritized.
The open question is whether labs will actually widen the door. OpenAI's restrictions on GPT-5.6-Cyber are tight, and the coalition's demand for inclusive eligibility rules cuts against how frontier access usually works. The initiative's backers are betting that pressure from 40-plus organizations — and the threat data behind it — changes the math.
