KINETIC-369 / HSK-01
An AI kill switch for the last inch of actuation
HSK-01 is a hardware AI safety interlock designed to interrupt a defined actuator power path when a hardwired threshold is crossed. “Kill switch” describes that physical interruption—not a way to shut down an AI model or contain every downstream risk.
TRL 4 — component validation in a laboratory environment. No certified product or completed conformity assessment exists.
Claim / Mechanism / Boundary
A defined hardware interrupt, not a promise of general AI containment
01 / Sense the boundary
Discrete, unprogrammable 74HC CMOS logic evaluates a bounded physical envelope at the digital-to-physical translation point.
02 / Break the path
When the wired threshold is violated, the HSK-01 architecture is designed to sever the protected actuator's power path in under 200 microseconds.
03 / Bound the claim
Protection applies to the instrumented path. Digital-only harm, unprotected actuators, lying sensors, in-envelope commands, and the human restore path remain outside that guarantee.
Hazardous condition sensed
Hardwired threshold crossed
Power path opened
Protected load de-energized
Conceptual path only. The interrupt applies to an instrumented actuator and depends on system integration, sensor integrity, and the restore path.
Technical comparison
HSK-01 hardware interlock vs. software-only controls
These are complementary controls, not interchangeable protections. Behavior and response depend on the implementation and the protected system.
Scroll table to compare →
| Design question | HSK-01 hardware interlock | Software-based control |
|---|---|---|
| Decision point | Hardwired threshold on a defined actuation path | Policy or logic evaluated in running software |
| Intervention | Opens the protected actuator power path when its threshold is crossed | Issues a stop, inhibit, or shutdown command through a software-controlled path |
| Software dependency | No processor, firmware, or network interface in the HSK-01 safety path | Depends on the controller, software state, and available command path |
| Coverage | Only the instrumented path and its wired envelope | Can evaluate broader digital states, subject to its implementation and authority |
TRL 4 architecture; this is not a certified product or a claim of system-level qualification.
Technical FAQ
AI kill switch: what it means here
- What is an AI kill switch?
- It is a way to stop a specified AI-driven action. Here, the term means interrupting the physical power path of a protected actuator—not turning off an AI model, network, or every device it can influence.
- How does HSK-01 work as a hardware interlock?
- Discrete 74HC CMOS logic evaluates a wired threshold. When that boundary is crossed, the HSK-01 design opens the defined actuator power path in under 200 microseconds. It must be integrated in series with the path it is intended to protect.
- How is this different from a software-only shutdown?
- A software shutdown sends a stop instruction through a controller or network. HSK-01 is designed to make the final interruption in hardware on the wired path, independently of the model's willingness to stop. It does not address digital-only harm, unprotected actuators, lying sensors, or in-envelope commands; software controls remain useful alongside it.
Technical review
Evaluate HSK-01 against your actuation path
Send your platform and application focus through the existing secure request form. A mutual NDA and technical scoping precede controlled release of the Technical Data Package. The TRL 4 architecture can be evaluated for direct military testbed integration through a Prototype OTA or, where qualifying prior work and acquisition determinations support it, Phase III SBIR/STTR. Neither pathway is an automatic award; integration and verification remain platform-specific.