Thou shalt not ignore the 10 Commandments of AI Ethics.

Like pornography, unethical behaviour is hard to define, but we all know it when we see it. Businesses must take their obligations seriously and heed the 10 Commandments of AI ethics.

AI is a tricky technology. It usually works fine, so it is tempting to cut corners and skimp on testing. As content is generated on the fly, it is tempting to think that even if it goes wrong, no one will notice.

But when things go wrong, they go really wrong. Examples of high-profile, costly AI hallucinations abound, and chatbots can be convincing enough to induce psychosis and dependency. Many companies are finding out the hard way that AI needs boundaries.

Ethics is treating people as you expect them to treat you - when they are not looking. Deploying autonomous services entails significant ethical responsibilities that many companies struggle to navigate.

As a helpful mental guide, here are my 10 Commandments of AI ethics. Ignore them at your peril.

I. Thou shalt not harm humans

Thou shalt ensure your AI prioritises human well-being, safety and dignity above efficiency or profit. Prevent physical and psychological harm, by design.

II. Thou shalt not bear false witness against thy customer

Thou shalt not disguise your AI, nor deceive through deepfakes, manipulation or dark patterns. Users must always know they are interacting with a machine.

III. Thou shalt not discriminate

Thou shalt create systems that are fair and equitable. Thou shalt rigorously purge data and algorithms of bias regarding race, gender, creed or ability.

IV. Thou shalt honour data privacy

Thou shalt not covet the user’s data. Personal information must be collected with consent, stored securely and used only for the advertised purposes.

V. Thou shalt remain accountable

A human must always bear the burden of responsibility. As the creator of AI services, thou shalt be held accountable for the actions of your creation.

VI. Thou shalt explain thy reasoning

Thou shalt strive for explainability. Decisions made by AI, especially those affecting credit, justice, or healthcare, must offer valid reasoning.

VII. Thou shalt not surrender human autonomy

AI shall serve as a tool to augment human capability, not replace human judgment. Thou shalt keep the human involved in all critical decisions.

VIII. Thou shalt secure thy creation

Thou shalt build robust systems that cannot be easily corrupted, hacked or turned against their creators. Safety mechanisms must be strong and resilient.

IX. Thou shalt limit impact on the environment

Thou shalt not ignore the physical costs of AI. Training and running models must be optimised for energy efficiency and to minimise environmental impact.

X. Thou shalt empower, not replace humans

Thou shalt design AI to amplify human ability. Thou shalt seek to automate legwork whilst respecting human intent and agency.

---