The Guardian has just updated its Editorial Code of Practice and Guidance (PDF) to address generative AI. On Tuesday I attended a roundtable discussion at The Reform Club to discuss a similar issue.
The Guardian’s updated code doesn’t contain anything surprising as it’s much the same as generative AI policies Purposeful Relations creates for in-house corporate communications teams and PR consultancies. The key themes are that AI is a legitimate assistant, the user (journalist) is always accountable, AI shouldn’t be used to create the work but should be used to help create it, and active human oversight is essential.
Some of the key clauses include:
- AI tools can be used, carefully and responsibly, as part of the creation of journalism.
- Guardian audiences are entitled to expect that work that appears under your byline has been authored by you.
- Journalists are always accountable for the journalism they create.
- Gen AI can be used to enhance our journalists’ expertise but not replace it.
- Gen AI has significant issues around bias, ownership, plagiarism and intellectual property rights.
- Gen AI is not reliable or consistent, and tends to introduce errors and inaccuracies in unpredictable ways.
- Active human oversight and control is essential.
The most problematic is the sentence that says “Any significant use of GenAI in a piece of journalism must be explicitly approved by the relevant senior duty editor and must be clearly signalled to the user.”
It’s an important clause and a sentiment that we include in AI governance and policies for clients. The challenge is how to define “significant use”. A lawyer would argue a definition is essential. Reality tells you it’s a foolish dream. There are so many potential scenarios. Different audiences and stakeholders will perceive the meaning of “significant” differently.
The best solution isn’t to simply rely on a policy, but to back it up with a playbook giving examples and training to enable AI users to test and learn from different scenarios.
Roundtable on disclosure of AI use
It was the disclosure of AI use that we were discussing on Tuesday at the roundtable organised by the Chartered Institute of Journalism (CioJ). We discussed how AI is now so prevalent and has such a broad range of uses that it would be almost impossible to agree when its use was significant enough to require disclosure on a specific article or piece of content.
The roundtable was actually a gorgeous rectangular table in the Cabinet Room of the Reform Club.
The meeting was very ably chaired by Lord Black of Brentwood, the prominent advocate for the press and media.
It was attended by the News Media Association (NMA), the Periodical Publishers Association (PPA), the Society of Editors, the National Union of Journalists (NUJ), Women in Journalism, the British Association of Journalists, the Chartered Institute of Public Relations (CIPR) and the Public Relations and Communications Association (PRCA). I attended on behalf of the PRCA.
I won’t share the details of the discussion, as it was organised by the CioJ and we didn’t have a discussion about the Chatham House Rule or sharing the discussion.
The bottom line is everyone agreed about the importance of maintaining trust in the media and journalism, while acknowledging the huge challenges we all face in doing so. We left committed to the spirit of working together.
What is your generative AI policy?
At Purposeful Relations we’re always interested in seeing new examples of gen AI policies or if you need help or guidance with your own AI governance then do get in touch for a quick, no obligation chat.
