Ethical GenAI and the Society of Authors
This is not a blog post I wanted to write.
A couple of days ago my trade body, the UK Society of Authors, released a Charter of Ethical AI Use in the Creative Industries, with a raft of co-signatories including the Australian Society of Authors, The New Zealand Society of Authors Te Puni Kaituhi o Aotearoa (PEN NZ), and the Writers’ Union of Canada. As far as I can see, none of these authors’ bodies consulted their memberships on this document. They just signed it in our names.
You can read the Charter in full here. https://societyofauthors.org/charter-of-ethical-ai-use-in-the-creative-industries/ Its stated aim is to offer ethical standards and a voluntary code for generative AI use in the creative industries, and the summary points for signatories are as follows:
By signing up to this Charter your organisation commits to:
Promote responsible and ethical use of generative AI services: balancing the technological benefits with the environmental, social, societal, economic, democratic and ethical impacts for humans and the planet.
Comply with the evolving best practice recommendations on AI ethics and governance: with respect to UNESCO’s 10 core principles of responsible AI and also the CLEAR framework (outlined in the Brave New World report) which has been adapted for use with this Charter.
Where AI is used, it is used for good: to widen access, inclusion and opportunity and to increase discoverability, helping more creators, from more backgrounds, reach wider audiences, and help to make creative work more accessible to all.
Work mindfully in relation to impact: on human jobs, livelihoods and the environment, investing in people’s skills so that they can share in the benefits.
Which all sounds lovely, apart from the bit where it concedes by its very existence that ethical use of GenAI is possible and to be expected, asserts it can be used for good, offers publishers a way to virtue-wash their GenAI use, and doesn’t set any sort of measurable goal or required action.
This was greeted with howling rage on Bluesky, including a ratio the like of which I have not seen since the glory days of Twitter. The SoA followed this up with a blog post from the CEO which began “Yesterday, we shared our Charter of Ethical AI Use in the Creative Industries on social media, and I am sorry that we clearly failed to communicate its purpose.” (For the record, starting a defence with “sorry you can’t read good” works better when your members aren’t all word people. I’m just saying.)
In the interests of ever getting some work done again, I’m just going to copy paste my response below.
Dear Anna Ganley
I am writing in response to the update yesterday.
You start by saying "We understand generative AI to be harmful to the creative industries, to individual writers’ livelihoods ... We are clear that there can never be any truly ethical use of generative AI because of the way LLMs have been trained on copyright works without permission or payment; the detrimental environmental impacts; and the harmful effect it has on disempowered communities". I absolutely agree. Where we part company is that your proposed Charter explicitly states that ethical AI use is possible.
I understand that you regard this as realpolitik, a 'statement of intent', a framework for how things should be and what publishers should aspire to. I think it's jawdroppingly naive.
Publishers who are using AI in the hope of cutting staff numbers and reducing spend on creatives already know they are putting people out of work. That's literally the point.
Anyone who cares to read the news already knows about the harms AI is doing to the environment, to workers, to people near data centres, to democracy. If they are using it it's because they don't care.
This offers the SoA ethical imprimatur to anyone who wants to claim it. Without checks, without safeguards, without qualification or inspections or consequences, you're proposing to let publishers slap a "We're Ethical!" badge on their website and call it a day. Talk about devaluing your brand.
No publisher is going to read this Charter and think "Wow, I should find a way to retrain the four copy editors and six marketing people I'm going to sack!" or "Gosh, I'll invest in a proprietary model with a fully paid-for training set!" or "I was going to use AI to mimic an author's voice or artist's style, rather than pay them, but now I realise that's bad." Those things will not happen. You know they won't happen. This Charter simply offers those publishers a fig leaf for their GenAI use, whereby they can claim lofty aspirations and carry on as before. ("Where viable alternatives exist, we will favour services from developers who license their training data"-- come on. Anyone can sign this and then explain sadly that the alternatives just aren't viable.)
Again, I understand your aim is to set a high behaviour standard. But voluntary codes of practice depend on the majority of people involved being ethical in the first place. Since GenAI is, as you acknowledge, fundamentally unethical, it is naive in the extreme to expect its keen adopters to have a Damascene conversion.
They need to be forced, which means we need consequences. We need the SoA facilitating a 'name and shame' for publishers using GenAI's stolen work to replace creatives, or to push slop into the workflow (which puts an additional burden on authors who have to rewrite the slop copy and fix the slop edit). We need a strong united voice saying we don't accept this. We need our union pushing for a class action suit against the book thieves, and drawing a hard line that says we must be made whole before there's any talk of publishers ethically using LLMs. We need to give publishers a really hard time on this. I'm sure the last two days have made you aware that authors are capable of giving people a hard time.
The very existence of this Charter constitutes authors' bodies agreeing that "ethical" GenAI use is possible and acceptable. This cedes an astonishing amount of ground, and it is unconscionable that the Society of Authors should have decided to capitulate to the interests of AI companies and publishers without even consulting its members. You cannot serve two masters.
I remain convinced this is a gross misstep, and one you had no right to take in the names of your members, and as such my resignation stands.
Yours
The SoA is absolutely right to say there should be transparency over whether, how and where generative AI is being used in publishing. But transparency isn’t an end in itself: what good is that? We need to see who’s using it and then call them out. Where a publisher is using LLMs developed on what Microsoft director of applied science Brent Hecht called the “largest theft of labor in human history”, and doing so in an effort to increase their own profits at the expense of both creatives and the publishing staffers who ought to support them, we need our union to hold them to account. Not to give the publishers a fig leaf of claiming they’re “working mindfully” or that they will definitely use an ethical model “where viable alternatives exist”.
I do agree there should be an ethical framework enforced by legislation. A list of specific actionable conditions publishers must demonstrate they are actually meeting, rather than mindfully considering, would be a good start. I also get the SoA has no power to enforce anything on publishers, still less legislators. It does, however, have 12,000 members in the UK, most of them robbed and enraged, whose anger could be used to make publishers question if the AI game is really worth the candle. (If, for example, authors could report receiving slop “work” from publishers, so that it became necessaryfor publishers to check the slop before sending it and fix it themselves instead of dumping that work on authors, it might quickly come to seem less of an amazing time-saving device. Or if they held a list of publishers and agents who have committed to not using it, rather than using it ‘ethically’. Or, as noted, if the SoA spearheaded the class action against the AI companies we’ve all been asking for.)
I am very sure the people behind this Charter, many of whom are busy authors putting in a lot of work on a voluntary basis to help their fellows, have done so with all sincerity. I just think the decision is wrong. And a large part of the reason I think the SoA has reached this wrong conclusion is a reluctance to piss off the publishing behemoths as much as is required. It is very hard for even bestselling authors to call out their publishers, the people who put their books on the shelves, for fear that their books will stop being on the shelves. I am in no way immune to this fear.
But that’s why we need solidarity. That’s what unions are for. And in practice, mine has just decided to hand the other side a huge advantage without giving me a chance to object.
Ultimately, whether you think this is sensible realpolitik or an unforced capitulation, the Society of Authors should not have ceded this ground without consulting its members on the direction of travel. And I’ve resigned because no publisher wearing an “I’m an Ethical GenAI user!” badge as they burn the planet to cut costs on editors, translators, illustrators, designers, marketers, voice actors, and writers will be doing so in my name.