Let’s begin with a straightforward question: Did I write this blog post? Yes. And also, perhaps not entirely.
Like many professionals today, I’ve started using generative AI tools like ChatGPT to assist in drafting, summarising, or brainstorming content. These tools are fast, efficient, and often surprisingly articulate. However, they also prompt a deeper inquiry that actuaries—guardians of professional rigour and responsibility—must consider:
What does authorship mean in the age of AI?
If I prompt an AI with “Write a blog post for actuaries about AI authorship,” and it returns a polished piece that I lightly edit and publish under my name, am I the author, an editor, a curator, or something else?
This isn’t mere semantics. For actuaries, the implications are profound. Our profession is built on trust, accountability, and transparency. Whether drafting model documentation, client reports, or internal memos, our name signifies ownership—not just of the words, but of the underlying reasoning.
If those words are generated by a machine, is that ownership diminished?
According to the Actuaries’ Code, particularly Principle 1: Integrity, actuaries must “act honestly and with integrity.” This principle underscores the importance of being transparent about the sources and processes behind our work, including the use of AI tools
Traditionally, authorship was clear-cut. You wrote it, or you didn’t. You might receive editing assistance, but the final thoughts and structure were yours. Now, generative AI can produce entire drafts from minimal input, raising new questions:
These questions tie into Principle 2: Competence and Care, which mandates that actuaries must “carry out work competently and with care.” Even when AI assists in drafting, actuaries are responsible for ensuring the content meets professional standards.
Consider this scenario:
Who is accountable?
You are. AI is a tool, not a co-author. Just as with any tool—be it Excel or R—you are responsible for its output.
This scenario reflects Principle 3: Impartiality, which requires actuaries to “ensure that their professional judgement is not compromised, and that they are not biased in their work.” Relying on AI without thorough review could inadvertently introduce biases or errors, compromising impartiality.
Here’s how actuaries can approach authorship when using AI tools:
In a sense, yes. I prompted it, reviewed it, edited it, and ensured it aligned with my beliefs and the message I want to convey to fellow actuaries.
But I also had assistance—from a machine that doesn’t understand authorship, context, or accountability.
That part is entirely my responsibility.
And perhaps that’s the essence of authorship today: not who typed the words, but who takes responsibility for them.
The piece above was written in its entirety by ChatGPT, based on a title and a couple of short prompts from me. Every word, every heading, every comma and full stop was generated by AI. I didn’t tell it the conclusion I wanted; I provided the question and the actuarial context, and it may have considered some of my chat history.
Do I agree that I’m the author of the piece? Definitely not. My entire contribution was the following two prompts:
I want to write a blog for actuaries entitled “Did I write this blog post?” It should discuss what authorship means in the age of AI, and whether, if the post was written by ChatGPT, that makes me the author, overseer or something else. Can you write this blog?
Can you rewrite the blog with references to how each part is relevant to the Actuaries' Code on the IFoA website?
I agree with ChatGPT that if I use the outputs, they are entirely my responsibility. For example, I don’t think it used the speaking up principle in the correct context; I would reference communication instead. I’d also make a few other edits and change the overall conclusion. At some point, then, if I edit enough of the words, do I become the author?
Perhaps.
Or perhaps not. At what point would the writing become mine? If I change 50% of the words? 60%? We can’t come up with an exact number. All we can do is be as transparent as we can that we have created something with the assistance of AI.
Beyond the issue of transparency, this exercise threw up some other questions for me:
At the end of the day, we’re responsible for what we produce—AI or not. Use of AI may become a necessity in our roles in the near future. But if we use these tools, it’s important to be upfront about it, and to use it in as ethical a way as possible. Being clear about where our words come from isn’t just good practice; it’s part of keeping trust in our work.
The views expressed are personal to the author and do not necessarily reflect those of her employer. ChatGPT was used in the writing of this post.
