
AgentBlackBox – a policy gate and tamper-evident audit log for AI agents
AgentBlackBox records what an AI agent saw, decided, requested, changed, blocked, and executed — and who approved it — into a tamper-evident, cryptographically chained audit trail you can prove to a customer, auditor, o…
以下正文同步自 Hacker News 最新,版权归原站所有,已转换为易读排版。
AgentBlackBox records what an AI agent saw, decided, requested, changed, blocked, and executed — and who approved it — into a tamper-evident, cryptographically chained audit trail you can prove to a customer, auditor, or regulator.
AgentBlackBox sits between your AI agents and the systems they touch and records a complete, ordered, tamper-evident log of their behavior and the human approvals around it. Each work unit (a task an agent is handling) gets a timeline of events, a policy gate, optional human sign-off, and a one-click audit report.
Autonomous agents now read inboxes, move money, change records, and message customers. When something goes wrong — or a customer, auditor, or regulator asks “what exactly did the AI do, and who approved it?” — you need a provable answer. A tamper-evident hash chain makes the record trustworthy: nothing can be quietly edited after the fact without detection.
Every event is SHA-256 hashed and linked to the previous one (prevHash → eventHash) from a GENESIS root. Any edit, insert, or delete breaks the chain and is instantly detectable.
Records what an agent observed, decided, requested, executed, blocked, or errored on — in order, with payloads.
High-risk actions gate on a human decision. Approvals and denials are themselves recorded as chained events.
Simulate any proposed action and get ALLOW, REQUIRE_APPROVAL, or BLOCK with the matched rules and a recommendation.
One-click, investor-ready audit report per work unit with executive summary, full event log, Print / Save as PDF, and JSON export.
External AI agents POST events to /api/events with an API key. Mock connectors (Gmail, Slack, Drive, Browser) show the future surface.
正文由 探How 从来源站点 RSS 同步,内容未经改写;遇到排版缺失或需要图片、视频时请以原文为准。