𝘿𝙤𝙚𝙨 𝙨𝙩𝙖𝙣𝙙𝙖𝙧𝙙𝙞𝙨𝙖𝙩𝙞𝙤𝙣 𝙠𝙞𝙡𝙡 𝙘𝙤𝙢𝙥𝙚𝙩𝙞𝙩𝙞𝙫𝙚 𝙖𝙙𝙫𝙖𝙣𝙩𝙖𝙜𝙚—𝙤𝙧 𝙢𝙖𝙠𝙚 𝙞𝙩 𝙨𝙩𝙧𝙤𝙣𝙜𝙚𝙧?
That tension was front and centre at the 𝗦𝘁𝗮𝗻𝗱𝗮𝗿𝗱𝗶𝘀𝗮𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝘁𝗵𝗲 𝗣𝗮𝘁𝗵 𝘁𝗼 𝗥𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗹𝗲 𝗔𝗜 panel at the Defence AI Research Network (DAIRNet) Defense AI and Industry Symposium (part of ADM Defence Summit). Here are some of the messages I was trying to deliver:

𝟭. 𝗦𝘁𝗮𝗻𝗱𝗮𝗿𝗱𝘀 𝗮𝗿𝗲 𝘁𝗵𝗲 𝗳𝗹𝗼𝗼𝗿, 𝗻𝗼𝘁 𝘁𝗵𝗲 𝗰𝗲𝗶𝗹𝗶𝗻𝗴.
They set the minimum bar, not the innovation boundary. The real edge lies in how well you implement them. Most standards operate at a high level—meaning there’s plenty of space for technical and operational innovation. And if you get smart about how you apply them, you can actually cut compliance cost significantly, turning a peceived burden into a strategic advantage.

𝟮. 𝗗𝗼𝗻’𝘁 𝗯𝗼𝗶𝗹 𝘁𝗵𝗲 𝗼𝗰𝗲𝗮𝗻—𝗳𝗶𝗹𝘁𝗲𝗿 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝘁𝗹𝘆.
Standards are designed for wide applicability, but any single use casem especially in defence, only needs a fraction of what’s listed. With the right lens, you can narrow hundreds of risk controls and best practices down to a manageable shortlist. CSIRO’s Data61 is actively developing tools, like structured question banks and reference architectures, that do exactly this: strip away the noise and surface what matters.

𝟯. 𝗛𝘂𝗺𝗮𝗻 𝗼𝘃𝗲𝗿𝘀𝗶𝗴𝗵𝘁 𝗶𝘀 𝘁𝗵𝗲 𝗵𝗮𝗿𝗱𝗲𝘀𝘁 𝗽𝗿𝗼𝗯𝗹𝗲𝗺.
Not whether humans are involved, but how. Oversight must be meaningful, well-timed, and anchored in the operational flow. Otherwise, it’s just theatre.

And a final thought related to human oversight. There was a recent study linking Alzheimer’s risk to different driving professions. The lowest? Emergency vehicle drivers—constantly making judgment calls. The highest? Airline pilots—operating extremely complex systems, yet heavily automated. A reminder that automation without cognitive engagement erodes human capability—and why human oversight in AI must be more than a checkbox.

For more on human oversight, see https://www.linkedin.com/feed/update/urn:li:activity:7343742258036281344/