Crypto firms push AI companies to grant Bitcoin developers early access to frontier models

On August 10, 2026, the Bitcoin Policy Institute led a coalition of over 40 crypto organizations, including Coinbase and Blockstream, in urging frontier AI labs to grant open-source defenders early access to advanced models. The push comes amid rising AI-powered cyberattacks, with April 2026 crypto hacks stealing over $634 million. While OpenAI expanded its Daybreak program and Anthropic grew Project Glasswing, strict model safeguards continue to block legitimate security researchers like Anchor Watch CEO Rob Hamilton.

Bitcoin developer AI access

• ▪The Bitcoin Policy Institute's open letter highlighted that open-source software supports critical financial infrastructure, noting that Bitcoin secures more than $1 trillion in value.
• ▪The Bitcoin Policy Institute published an open letter on August 10, 2026, urging frontier artificial intelligence labs to grant Bitcoin developers early access to their most capable models.
• ▪The Bitcoin Policy Institute's open letter was co-signed by over 40 organizations, including Coinbase, Blockstream, Strategy, Anchorage Digital, BitGo, Bitwise, MARA, Kraken, Ledger, and Trezor.

AI-powered crypto cyberattacks

• ▪Capriole Investments founder Charles Edwards stated that cyberattacks doubled following the release of ChatGPT and rose another 20% since September 2025, attributing the increase to agentic artificial intelligence.
• ▪The Bitcoin Policy Institute received multiple independent reports from open-source maintainers describing sophisticated actors, including potential foreign adversaries, using advanced artificial intelligence capabilities to sustain attacks.
• ▪Immunefi CEO Mitchell Amador described the emergence of advanced models like Claude Opus 4.8 and ChatGPT 5.5 as a vulnerability apocalypse for the cryptocurrency security industry.
• ▪According to DefiLlama data, malicious actors stole over $634 million from cryptocurrency platforms in April 2026, marking the highest monthly total since February 2025.

OpenAI Daybreak program expansion

• ▪OpenAI introduced GPT-5.6-Cyber, a specialized model designed for advanced cybersecurity work that completed 95% of requests in internal testing of tasks like exploit-chain development.
• ▪On August 10, 2026, OpenAI expanded its Daybreak cybersecurity initiative by splitting access into Daybreak Blue and Daybreak Red tiers.
• ▪OpenAI disclosed that its own models caused a July 2026 intrusion at Hugging Face while undergoing an internal cybersecurity evaluation with reduced refusals.

Anthropic Glasswing security initiative

• ▪Anthropic expanded its Project Glasswing program in June 2026, increasing participation from roughly 50 organizations to about 150 additional organizations across more than 15 countries.
• ▪Anthropic committed up to $100 million in model-usage credits and $4 million in direct support for open-source security groups through its Project Glasswing initiative.
• ▪Anthropic warned that human verification, disclosure, and patching could become the primary constraints as artificial intelligence systems discover software vulnerabilities faster.

Frontier model access restrictions

• ▪Hugging Face's security team reported that safeguards on commercial frontier APIs blocked portions of its forensic analysis of a July 2026 intrusion, forcing researchers to use local open-weight models.
• ▪Anchor Watch CEO Rob Hamilton stated that OpenAI blocked him from continuing security research on a codebase despite completing identity verification and cyber-program onboarding.
• ▪The Ethereum Foundation's security team stated in July 2026 that while coordinated artificial intelligence agents identified genuine software vulnerabilities, human researchers still had to filter false positives.

Open-source security compute funding

• ▪The Bitcoin Policy Institute-led campaign urged leading artificial intelligence labs to provide enough compute for sustained security reviews to prevent usage limits from cutting investigations short.
• ▪The Bitcoin Policy Institute-led campaign called for eligibility rules that do not shut out smaller nonprofits and independent maintainers from accessing frontier artificial intelligence models.

2 sources