We build hardware-accelerated ethics engines that let autonomous agents obey democratically governed norms in microseconds—not seconds.
Ethical Finite Machines starts from a simple premise: if we want autonomous systems to obey democratic norms, ethics can’t live as a slow, opaque afterthought. It needs to be a first-class component of the control loop, with bounded latency and transparent governance.
Ethical configurations are expressed as profiles: stakeholder weights, ethical dimensions, principlism layers, and hard vetoes. Profiles can be debated, versioned, and adopted by regulators, communities, or operators.
Rich EthicalFacts from perception and planning are compiled into compact EthicsFrames — fixed-width encodings of distance, risk, rights, vulnerable presence, and action type that can be evaluated in a handful of cycles.
FPGA/ASIC modules implement profile-aware hard vetoes and normative scoring in microseconds, so reflex and tactical decisions can stay within ethics without slowing the machine down.
We’re building the bridge between democratically governed ethical frameworks and the hard constraints of real-world autonomy: compiler toolchains, hardware EM cores, and governance infrastructure for fleet-scale deployment.
High-level DEME-style profiles are compiled into hardware artefacts: EthicsFrame mappings, fixed-point weights, veto predicates, and lexicographic priority rules ready for synthesis.
Our Hardware Ethics Modules run alongside perception and planning stacks, evaluating multiple candidate actions per cycle with bounded latency and clear veto semantics.
Every ethics decision can be traced: which profile, which bitstream, which vetoes fired, and which options were considered. Oversight agents can spot bias, drift, or profile violations at fleet scale.
Anywhere autonomous systems interact with people in physical space, ethical behaviour isn’t optional. Our focus is on domains where decisions need to be both auditable and sub-millisecond.
Lane changes, braking, path selection around vulnerable road users, and triage between conflicting risks all demand real-time ethics, not offline policies.
Robots operating in warehouses, hospitals, and public spaces need predictable responses when near people, critical infrastructure, or restricted zones.
Regulators and standards bodies can experiment with DEME profiles and Hardware EM implementations before mandating them, using shared test suites and audit tooling.
We’re exploring collaborations with AV and robotics teams, regulators, and researchers who care about democratically governed, real-time ethics for autonomous systems.
