How can organisations integrate ‘responsible AI’ into their communication strategies, ensuring ethical standards, transparency and human oversight?
As a speaker and writer, I’m often asked how I maintain originality when using AI. But this isn’t a creative challenge; it’s an ethical one. In my world, responsible AI use requires a blend of transparency, training and what I call ‘confident tweaking’.
Responsible AI: Transparency
To prevent misunderstandings and build trust, my clients must know when I’ve used an AI tool or Large Language Model, so I’ve developed my own AI policy.
I disclose when and where an LLM is likely to be deployed, which specific LLM/s I use, and where the LLM stops and the human starts (that is, when I step in and take over and why I always do).
Responsible AI: Training
Everyone who works for, with and around me is trained to use AI responsibly, because it doesn’t come naturally. Our training clarifies:
- Data exposure – What confidential information, proprietary ideas, trade secrets or sensitive data should never be entered into an LLM?
- Privacy – How can we safeguard customer details and business strategies? How can we use anonymized or dummy data?
- Ownership – Who owns which pieces of generated content? Me (the service provider), the client or the LLM provider?
Responsible AI: Confident tweaking
Human communicators must be able to deftly massage LLM-generated output, to maintain the balance between automation, authenticity and nuance.
Part of becoming AI-literate is learning to pre-empt and remedy limitations and biases. For example, LLM copy might nail the grammar but fall flat on cultural sensitivity – and require confident human tweaking.
