Responsible AI stance
Six principles for marketing, written to decide real cases
A principle that cannot rule out a specific decision is a slogan. These are written to be applied to work in progress, including our own, and they are the standard our teaching is built on.
01
Disclosure is the default
When a person could reasonably want to know that what they are reading, seeing or speaking to was machine generated, they should be told. Our position is that disclosure belongs in the work itself rather than buried in a policy page, and that the burden of proof sits with the marketer who wants to stay silent, not with the audience asking. We teach disclosure as a craft problem: how to say it plainly, in the right place, without turning every piece of content into a legal notice.
02
A person is accountable for every published thing
Automation can produce the work. It cannot hold responsibility for it. Every piece of marketing that reaches an audience should have a named human who approved it and who would answer for it if it were wrong, misleading or offensive. This is not a brake on speed, it is the thing that makes speed survivable. Teams that cannot name that person for a given output have a governance gap, not a technology problem.
03
Consent is not a technicality
Data that a customer provided for one purpose does not become available for every purpose because a new capability made it useful. The reasonable expectation of the person who gave you the data is the standard, and it is a higher standard than the minimum a terms-of-service page can be argued to permit. Australian marketers should expect community expectations here to keep tightening, and should build on the assumption that what is barely defensible today will be plainly unacceptable soon.
04
Personalisation has a limit, and it is set by dignity
The technical ceiling on how precisely a person can be targeted is now far above the ethical one. Inference about health, financial distress, sexuality, immigration status or vulnerability should not be used to sell, whether or not the inference is accurate and whether or not the data was lawfully obtained. We treat the question “would this person be uncomfortable knowing how we reached them” as a real design constraint rather than a discussion for later.
05
Likeness and voice belong to people
Synthetic representation of a real person requires that person’s informed, specific and revocable permission, including for people whose image or voice your organisation has previously licensed for conventional use. A general release signed for a photo shoot is not permission to generate new performances. This applies to staff and to customers as much as to talent, and the Centre teaches it as a hard boundary rather than an area for commercial judgement.
06
Say what you do not know
Modelled measurement produces estimates, and estimates carry uncertainty. Presenting a modelled number to a client or a board with the same confidence as a counted one is a form of dishonesty, even when it is unintentional and even when the model is good. We teach marketers to state their uncertainty out loud, because a profession that overstates its certainty will eventually be believed less on everything, including the things it is right about.
These principles are reviewed annually with the membership and with the Responsible Practice Director, whose specific brief is to argue against them. Where a member believes the Centre has failed to meet its own standard, that is a matter for the board and we would rather hear it early.