Hi, I'm Steve Bullock from the University of Bristol and Bristol Flight Lab. This is a ten-minute introduction to my part of the live session: how to use the sensing technologies Justin and Devis are talking about safely, ethically and legally. It goes with the short reading and the capability audit canvas on the course page.

Devis's crash course takes you from satellites down to a drone over a herd. Add camera traps, acoustics, eDNA and the models, and it's an extraordinary toolkit. Every capability in it is also a responsibility. I'll look through three lenses: safe, ethical, legal.

Safety and compliance overlap, but they aren't the same. If you only tick boxes you can still hurt someone; if you only trust your own judgement you can still end up in serious trouble. And neither guarantees you're doing the right thing. Aim for the middle.

Use your institution's fieldwork risk assessment, and the course's own field safety plan. The steps are the same everywhere: hazards, harm, controls, plan B, review. For drones beyond the lowest-risk category, Europe and the UK use SORA, which splits ground risk from air risk. Even if you never write one, that split is a useful way to think about any flight.

When you choose controls, work from the top. Removing the hazard beats everything else; personal protective equipment is the last resort, not the first. The examples on the right are from drone and fieldwork planning.

Two rules every drone lab learns the hard way: use procedures and checklists, and never have the props on and the battery connected unless you're about to fly. The field lessons apply just as much to a camera-trap grid or an acoustic array: work with local people, do the paperwork early, plan rest, and expect to go back more than once.

Your risk assessment should include the animals: there's good evidence on drone disturbance and good practice for reducing it. And it should include people downstream of your outputs. Automated pipelines make mistakes at scale. Who checks?

Design choices decide what a system can do, and to whom: which sensors, how much autonomy, what data you keep, what you release. Every one of these is a real example, and conservation technology has been used against the communities who live where it's deployed. Sandbrook and colleagues' 2021 principles are the best single place to start.

Value-sensitive design treats ethics as part of design, not a form at the end. The values hierarchy turns an abstract value into a norm, and the norm into something you can build or test. Read it top-down to specify, or bottom-up to justify why a requirement exists.

Capability caution is the design habit that goes with it: an alternative to maximising every capability. Set "good enough" for the mission, then stop. These five principles are the questions the canvas asks you to work through for your own system.

Open data makes science better, but open isn't always fair or safe. FAIR and CARE work together. GBIF has good guidance on generalising sensitive species locations. And find your ethics committee early: approval can take months.

The UK and EU regulate by risk, not by why you fly, so these first qualifications are a real first step for research. The UK rules changed in January: you now need a Flyer ID from 100 grams. In the EU, use your national authority's official site; there are fake licence sites.

The US does distinguish by purpose. Training courses, Embry-Riddle's included, say if it isn't for fun you need Part 107, and for most people that's right. But since 2021 the law has counted university research as recreational, within all the recreational limits. It doesn't cover companies, agencies or NGOs, and many universities require Part 107 anyway.

Drones are where the law is most visible, but these apply to every sensor. Notice how many of them are about data: privacy, sensitive locations, licences. That's where the AI side of this course meets the law.

Three things before the live session, all linked from the course page. In the session you'll swap canvases with a partner and red-team each other's: find the worst plausible misuse of their system, then decide what you'd change in your own. See you there.