展示HN：我们申请了99项关于确定性人工智能治理的专利（现有技术与强化学习人类反馈的对比）
在过去几个月里，我们一直在致力于自主智能体治理的根本架构转变。目前行业标准几乎完全依赖于概率性对齐（如强化学习人类反馈、系统提示、宪法训练）。这种方法在被破解或上下文窗口溢出时会失效。统计性质并不是安全边界。
我们构建了一种替代方案：确定性策略门。在我们的架构中，语言模型（LLM）完全剥离了执行权力。它只能生成一个“意图载荷”。该载荷会被传递到一个进程隔离的、确定性的执行环境中，在那里它会根据一个经过加密哈希的约束矩阵（宪法）进行评估。如果违反了该矩阵，则会被阻止。每个决策都会被记录到一个Merkle树底层（GitTruth）中，以便形成不可篡改的审计轨迹。
我们于2026年1月10日提交了99项关于该架构的临时专利。至关重要的是，我们在专利声明中直接嵌入了严格的人道主义使用限制（和平机器授权），以确保这些知识产权不能被合法用于自主武器、大规模监控或剥削。
我撰写了一份关于该架构的详细分析，解释了为什么概率性安全是一条死胡同，以及我们在行业发布其框架之前提交专利的时间线：请在此阅读完整宣言：<a href="https://salvatoresystems.medium.com/the-death-of-probabilistic-safety-99-patents-for-deterministic-ai-governance-888e51a14a04" rel="nofollow">https://salvatoresystems.medium.com/the-death-of-probabilist...</a>
完整的专利注册信息可以在这里公开查阅：<a href="https://aos-patents.com" rel="nofollow">https://aos-patents.com</a>
我是创始人和唯一发明人。欢迎随时询问关于确定性架构、Merkle树状态持久性或将伦理直接嵌入专利声明中的知识产权策略的任何问题。
查看原文
For the last few months, we've been working on a fundamental architectural shift in how autonomous agents are governed. The current industry standard relies almost entirely on probabilistic alignment (RLHF, system prompts, constitutional training). It works until it's jailbroken or the context window overflows. A statistical disposition is not a security boundary.<p>We've built an alternative: Deterministic Policy Gates. In our architecture, the LLM is completely stripped of execution power. It can only generate an "intent payload." That payload is passed to a process-isolated, deterministic execution environment where it is evaluated against a cryptographically hashed constraint matrix (the constitution). If it violates the matrix, it is blocked. Every decision is then logged to a Merkle-tree substrate (GitTruth) for an immutable audit trail.<p>We filed 99 provisional patents on this architecture starting January 10, 2026. Crucially, we embedded strict humanitarian use restrictions directly into the patent claims themselves (The Peace Machine Mandate) so the IP cannot legally be used for autonomous weapons, mass surveillance, or exploitation.<p>I wrote a full breakdown of the architecture, why probabilistic safety is a dead end, and the timeline of how we filed this before the industry published their frameworks: Read the full manifesto here: <a href="https://salvatoresystems.medium.com/the-death-of-probabilistic-safety-99-patents-for-deterministic-ai-governance-888e51a14a04" rel="nofollow">https://salvatoresystems.medium.com/the-death-of-probabilist...</a><p>The full patent registry is public here: <a href="https://aos-patents.com" rel="nofollow">https://aos-patents.com</a><p>I'm the founder and solo inventor. Happy to answer any questions about the deterministic architecture, the Merkle-tree state persistence, or the IP strategy of embedding ethics directly into patent claims.