Niclas Braun — Research and Essays Independent work on AI safety, physical trust, verifiable authority and resilient autonomous systems. https://niclasbraun.com/ 2026-07-26T12:25:00+02:00 Niclas Braun https://niclasbraun.com/ AI Kill Switches Are Not a Control Architecture https://niclasbraun.com/thinking/ai-kill-switch-not-control-architecture 2026-07-26T12:25:00+02:00 2026-07-26T12:25:00+02:00 Why emergency shutdown is necessary but insufficient—and why enforceable AI safety must begin before capability becomes authority. Physical Trust at the Digital-Physical Boundary https://doi.org/10.5281/zenodo.21471429 2026-07-21T00:00:00+02:00 2026-07-26T12:15:00+02:00 An information-theoretic framework for verification where digital representations meet decision-relevant physical states. Verification as a Control Primitive for Frontier AI https://doi.org/10.5281/zenodo.21454914 2026-07-20T00:00:00+02:00 2026-07-26T12:15:00+02:00 A systems perspective on enforceable authority in high-capability environments. AI Safety Has an Enforcement Problem https://niclasbraun.com/thinking/ai-safety-has-an-enforcement-problem 2026-07-20T00:00:00+02:00 2026-07-20T00:00:00+02:00 Why increasingly capable AI systems need enforceable limits at the boundary between decision and action. AI Agent Authorization: Identity Is Not Permission https://niclasbraun.com/thinking/ai-agents-need-verifiable-authority 2026-07-19T00:00:00+02:00 2026-07-26T11:58:50+02:00 A practical model for bounded, contextual and revocable authority in AI-agent systems. Trust After AI https://niclasbraun.com/thinking/trust-after-ai 2026-07-19T00:00:00+02:00 2026-07-19T00:00:00+02:00 Why the physical world needs a verification layer when digital evidence can no longer be trusted by default.