作者认为，当前的AI系统，尤其是在公共部门，缺乏真正的问责制，因为它们的决策过程和解释是不可审计的。他们提出需要一个“主权底层”，其中包括加密执行和对AI所采取的每一个行动进行防篡改记录。这个底层将确保决策可以被质疑，并且系统的行为可以被验证，从而弥合合同协议与实际操作行为之间的差距。 AI
影响 真正的AI问责制依赖于可审计的、防篡改的记录，而不仅仅是模糊的解释，以确保决策可以被质疑。
排序理由 该集群由讨论AI问责制和需要可审计记录的观点文章组成，而不是像发布或资助这样的具体事件。
在 Mastodon — fosstodon.org 阅读 →
AI 生成摘要 · Google Gemini · 来自 6 个来源。 我们如何撰写摘要 →
Public Sector AI and the Decision a Citizen Can Challenge When the state automates a decision about a citizen, the test is not accuracy but accountability: can the person challenge it, and does the record survive scrutiny. We look at why most public sector AI fails that test, and…
The Agent That Spends Needs an Envelope, Not Just a Budget When an autonomous agent can move money, a soft budget is a suggestion and a hard envelope is a law. The difference is cryptographic enforcement and an auditable record, which is exactly what a sovereign substrate provide…
Explainability Without a Witness Is a Story You Cannot Check Model explanations are after-the-fact narratives that nobody can audit. Real accountability needs a tamper-evident record of the actual decision, sealed and signed at the moment it happened. https:// mickai.co.uk/articl…
Sovereign Compute Is a Lease Unless You Hold the Update Channel Owning the hardware is not the same as owning the compute. Whoever holds the update channel, the signing keys, and the record of what actually ran is the real sovereign. Here is how Mickai closes that gap. https:// m…
The Liability Gap the Contract Cannot Close A contract allocates liability on paper, but it cannot prove what actually happened. That gap between the clause and the conduct is where disputes live. Mickai closes it by sealing each action into a signed, anchored record. https:// mi…
The Kill Switch Nobody Can Prove They Pulled Automated systems refuse, halt and revoke in milliseconds, but most cannot prove who decided, on what authority, or that the stated reason is the real one. Mickai closes that accountability gap with local intelligence, a post-quantum-s…