When I started Mumsnet back in 2000, the forum largely consisted of me talking to myself through multiple other personas. I had no idea that nearly 25 years later, millions of other users would have added over six billion words, discussing everything from parking to geopolitics to the infamous “penis beaker” discussion about one user’s post-coital clean-up regimen (which generated so much traffic it brought the website to its knees for a fortnight).
And now, those six billion-plus words are at the heart of one of the big questions of our age – how do we balance AI progress with copyright protections for publishers?
Last year, Mumsnet was the first company to begin legal proceedings against tech giant OpenAI, creators of ChatGPT, in the UK. We became aware they were scraping our site in order to train their large language model (LLM) – which we believe is a breach of both our terms of use and copyright law.
When we suggested they might like to license our content, after some prevarication they told us that they were more interested in “less open” sources of data. Truth is, they’d already scraped it all – something that they’ve not denied in all our correspondence.
The irony is that we’d be happy to work with companies like OpenAI, and use the female-dominated conversation on Mumsnet to help tackle the misogyny baked into most AI models. We use OpenAI’s LLM in our own MumsGPT – a question tool that provides insights based on conversations between Mumsnetters. But the difference is that we licensed and paid to use that technology.
Mumsnet is in a better position than most when it comes to withstanding the fallout of AI chatbots. Much of our traffic comes to us direct, and though an LLM can regurgitate a Mumsnet-style answer to a question, it’s no match for the rich, empathetic and often hilarious conversations on site. When your baby won’t sleep, your standard AI rehash will never be as authentic as an answer from someone going through exactly the same thing as you at 3am.
Nor will AI ever be as funny or as brutally honest. They’ll certainly never provide the support that sees around 1,000 women a year helped to leave abusive relationships by other Mumsnetters.
But if the tech giants are allowed to steal from publishers with impunity, they will destroy many of those publishers. That’s why it’s so surprising this government seems determined to abandon our longstanding copyright law. Their proposals would mean publishers have to opt out of the scraping of their content, or have it taken without payment or consent – akin to saying it’s OK for someone to burgle your house unless you’ve put a sign up with their name on telling them not to.
This approach doesn’t just threaten those being scraped – it risks undermining the models as well. If it’s no longer financially viable for publishers to produce the content that is being scraped, and go bust, then eventually there will be nothing left for AI companies to train their models on.
The Government has made much of the “uncertainty” they say currently surrounds the UK’s copyright regime in relation to AI. But the truth is that our robust system is one of our greatest strengths. If all the Government wants to do is introduce more clarity, then why did it oppose the Kidron amendments introduced into the Data Bill, which would make the existing copyright regime work in the age of AI?
AI has the potential to advance human progress and improve our lives, and of course we want the UK to reap its benefits. But that doesn’t mean publishers should give away their content to AI giants for free. Tech firms don’t expect the electricity that powers their data centres to come without a cost – so why shouldn’t they compensate the creators who produce the content they train their models on?
The UK can lead the world in this field. But in order to do so, we need to build a sustainable ecosystem, where innovation and fairness coexist.
Justine Roberts is the founder and CEO of Mumsnet
Comments
Leave a Comment