Nyx Foundation and bitbank Sign Joint Research Deal on AI Agent Trading Safety
Nyx Foundation, a general incorporated association, signed a joint research agreement with bitbank, the SBI Group cryptocurrency exchange operator, to improve the safety of cryptocurrency trading by AI agents.
1 source 5m ago
Reporting from 1 source: ASCII.jp.
Why it matters
bitbank already released experimental software in July 2026 letting AI agents use exchange functions, and this agreement moves the question from whether agents can trade to how their delegated authority gets bounded and verified.
Key Facts
- Nyx Foundation signed a joint research agreement with bitbank, which operates the bitbank cryptocurrency exchange within the SBI Group.
- The joint research covers managing trading authority delegated to AI agents and preventing orders that exceed user-set conditions.
- Nyx's side of the work uses security technology and mathematical verification methods, including formal verification.
- bitbank released the experimental software bitbank-lab-mcp and bitbank-lab-cli in July 2026 for AI agents to use exchange functions.
- The research addresses cases where multiple orders, partial fills, or cancellations push past a user's cumulative limit.
The work combines bitbank's trading knowledge with Nyx's security technology and mathematical verification methods. The goal is a system mechanism that enforces user-set spending limits and trading conditions, rather than relying on instructions given to the AI agent itself.
The research targets a gap that simple order checks miss. A user who permits a cumulative total of 500,000 yen is not protected if each individual order is checked only against that ceiling, since several orders can add up past it. Partial fills and cancellations also require the system to track what has already been used against what may still be used.
Three technical subjects are named: defining how delegated authority is expressed and managed on the system, building a mechanism that blocks orders exceeding the set conditions, and verifying safety with a mathematical method called formal verification. bitbank's own experimental tools, bitbank-lab-mcp and bitbank-lab-cli, were released in July 2026 to test AI agent use of exchange functions.
Written with AI by Yomimono from the 1 cited source below, including Japanese-language reporting where cited. See our editorial policy.
Add Yomimono as a preferred source on Google