# Coinbase, Block and BitGo Ask Artificial-Intelligence Labs to Unlock Frontier Models for Bitcoin Security Work

More than 40 organizations argue that safety guardrails block defenders from the same code-analysis capability attackers already use without restriction.

- Published: 2026-08-13T05:53:29.271Z
- Canonical: https://polylog.news/crypto/2026-08-13/coinbase-block-and-bitgo-ask-artificial-intelligence-labs-to
- Publisher: Polylog (Crypto desk)
- Section: crypto
- Sources: [CoinDesk](https://www.coindesk.com/tech/2026/08/13/bitcoin-firms-ask-ai-labs-for-same-tools-attackers-already-have), [crypto.news](https://crypto.news/bitcoin-defenders-seek-frontier-ai-access-group-push/), [Cointelegraph](https://cointelegraph.com/news/crypto-companies-ai-firms-bitcoin-developers-access)

The Bitcoin Policy Institute and more than 40 companies and groups, among them Coinbase, Block, BitGo, Strategy, MARA, Galaxy, Brink and Trezor, [signed an open letter to frontier artificial-intelligence laboratories](https://www.coindesk.com/tech/2026/08/13/bitcoin-firms-ask-ai-labs-for-same-tools-attackers-already-have) asking for vetted access to the strongest models for defensive security research. The signatories say public safety guardrails on those systems refuse legitimate vulnerability work, so open-source maintainers fall back on weaker open-weight models while attackers face no equivalent limit.

The request is specific. It asks for early access to frontier cybersecurity models, enough computing capacity to run large analyses, secure research environments, and direct contact with laboratory security teams, so that vetted researchers can [review wallets, payment infrastructure and other open-source software](https://crypto.news/bitcoin-defenders-seek-frontier-ai-access-group-push/) before attackers find the same defects. The coalition points to more than $634 million stolen from crypto platforms in April 2026, citing DefiLlama data, and to the roughly $1 trillion of value secured by Bitcoin.

The asymmetry the letter describes is already visible. Aztec found a critical flaw in its own proving system through internal artificial-intelligence-assisted auditing, and volunteer reviewers have used similar tooling to surface long-dormant defects in widely used open-source cryptography libraries. Whether laboratories grant privileged access is a commercial and liability decision, not a technical one, and no laboratory has publicly [committed to the program](https://cointelegraph.com/news/crypto-companies-ai-firms-bitcoin-developers-access).

## What this means

Custody and exchange businesses carry the cost of every unpatched defect in the open-source software they run, so they are asking the model providers to take on part of that defensive work. If laboratories create a vetted-access tier, the largest, most institutionally connected firms get it first, and small independent maintainers stay on weaker tooling, which concentrates security capability alongside capital. If laboratories refuse, the disclosure rate stays uneven, with well-resourced attackers and a handful of internal teams running the strongest analysis.

## What to watch

- Whether any frontier laboratory announces a vetted defender program with named participants, which would show that a formal two-tier access model has arrived.
- Whether monthly theft totals tracked by DefiLlama fall in the quarters after any such program starts, the only real test of whether defender access changes outcomes.
- Whether independent Bitcoin Core contributors, rather than corporate signatories, are actually included, since exclusion would leave the software everyone depends on reviewed by a corporate subset.