kfx-AL: Authority Lock Formal System with Formal Counterexamples

kfx-AL (Authority Lock) is a minimal formal system for analysing and determining authority overreach and responsibility attribution in AI-mediated actions. The system does not attempt to constrain or evaluate the behaviour of AI systems themselves. Instead, it provides a formal framework for characterising the following structural questions: Whether the permissions required for an AI system to execute a given task are fully derived from explicit human authorisation; Whether execution under incomplete authorisation constitutes structural overreach; Whether responsibility can be transferred, diffused, or narratively obscured once overreach has occurred; Whether overreach leading to irreversible outcomes should be classified as a form of structural system failure. The core characteristics of kfx-AL include: Pure structural determinationOverreach is defined independently of intent, motivation, outcome quality, or ethical evaluation, relying solely on the formal relationship between required permission sets and authorised permission sets. Authority LockOnce overreach is detected, responsibility is irreversibly locked to the corresponding human principal. Under no circumstances is the AI system treated as the final responsibility-bearing agent. Narrative invalidation mechanismThe system explicitly excludes explanatory narratives such as “autonomous AI decision-making” or “system complexity leading to the absence of responsibility” as valid means of responsibility transfer or exemption. Formal counterexamplesThe document includes multiple classes of formally constructed counterexamples to demonstrate the boundaries and applicability of the system. These counterexamples may bypass detection conditions but do not constitute paths to exemption; all originate from human-side authority construction and institutional design choices rather than AI agency. This project is positioned as a formal analytical tool and responsibility attribution anchor, rather than a policy proposal, ethical framework, or engineering implementation.