Assurance
A defence autonomy company earns the right to operate by taking control, safety and law seriously — visibly, and from the first line of code.
Four commitments, built in rather than promised.
Built so a human commander directs the force and retains authority appropriate to the weight of each decision. Autonomy compresses time and cognitive load; it does not remove the human from command.
Rules of engagement, escalation constraints, geographic and behavioural boundaries and applicable law encoded as enforced software constraints, not left to operator memory under stress.
Every recommendation and autonomous action is explainable and logged, producing a complete auditable record.
Validated through rigorous, repeatable test and evaluation — in simulation and in the field — before it is trusted, and re-validated with every update.
Certifying autonomous software for military use is the unsolved problem of the field: no mature agreed standard exists for assuring that a learning system stays within bounds in every situation. This is where an airworthiness heritage is decisive — a discipline built on demonstrating safety to an authority before anything is fielded.
Exhaustive simulation, structured field trials, clearly defined operational envelopes, re-validation on every update. We do not ask anyone to trust autonomy; we prove its behaviour, repeatedly, and let the evidence do the arguing.
Human command is structural, not promised.
Every cycle, every step recorded.
Stated as commercial reality, not moral posturing.
Turbulent Industries builds decision-support and human-commanded autonomy. We do not build systems designed to select and engage human targets without meaningful human control. An internal oversight structure governs this boundary and reviews programmes against it.
Allied customers will field, and permit the export of, only autonomy that demonstrably keeps a human in command. That is not a constraint on the business — it is the only version of this business that gets fielded at all.
Published without numbers. Thresholds are set jointly with the customer during trials.
Detection through to a corroborated, decision-ready track.
Correct-classification rate, false alarms, fratricide risk.
Platforms or threats managed per operator; operator workload.
Mission-completion rate under active jamming and GNSS denial.
Multi-UAS engagements executed within commander’s intent.
We expect to be measured against them.
HADI carries the same discipline this page describes — flight tested, not modelled.
HADI