Weekly take: IP concern and excitement – one year of ChatGPT
As ChatGPT celebrates its first birthday, we are still grappling with a multitude of IP concerns
My sister recently texted our family WhatsApp group to share the lyrics to a song she had asked ChatGPT to write about herself.
She had asked the tool to base the song on a music therapist who likes swimming and plays the piano – a little insight into my sister’s life there.
The lyrics were reasonably good, if a little trite.
I couldn’t see anything that had been pinched from elsewhere so she probably would have been safe from any intellectual property infringement claims had she shared it to a wider audience.
I hazard a guess many people have had similar fun with ChatGPT over the past year.
In fact, last Thursday, November 30, marked the tool’s first birthday.
Its arrival back in November 2022 was met with an understandable buzz.
But it also attracted plenty of headlines (and still does) over its alleged citations of fake news, non-existent court judgments, and misinformation.
In an article for Managing IP earlier this year, lawyers at law firm Linklaters gave an apt example.
They revealed that ChatGPT, in an early test of its capabilities, had told them that the late Lord Denning, who had been a judge at the England and Wales Court of Appeal, had been arrested for shoplifting in 1957.
The claim has not been substantiated anywhere else and so appears to have been entirely made up.
Of course, you cannot defame the dead, but had similar claims been put out about a living person then a defamation lawsuit would almost certainly have been on the cards.
ChatGPT’s reliability may have got better (and will doubtless continue to improve), but while it is arguably better at rooting out fake news and defamatory remarks, we are seemingly not much clearer on its impact on IP law than we were 12 months ago.
Race to the bottom
One of the foremost concerns to have developed over the past year is the potential for copyright infringement, particularly regarding the training of the ChatGPT tool.
Some jurisdictions, such as the EU and the US, allow for text and data mining (TDM) under certain circumstances. The EU’s TDM exception was introduced in the Copyright Directive, which came into force in 2019, while the US’s fair use doctrine permits the mining of some material without prior permission.
The UK proposed TDM legislation but it was binned following a backlash from rights owners earlier this year. The government has since outlined its desire to develop a code of practice related to AI, and that is expected by early next year.
In fact, governments the world over are racing to try to update their laws to create the best AI regimes.
The UK has been particularly bullish in this regard with several recent prime ministers (I’ve lost count of how many the country has had in the past few years) expressing their desire for the UK to become an AI leader.
The UK, according to its lawmakers, must compete with other jurisdictions, such as the US and the EU in “setting the pace” for AI regulation, while Russia has previously claimed that those that lead in AI will be “rulers of the world”.
According to the aforementioned Linklaters article, this shared desire to be AI leaders could risk creating a “race to the bottom” – at least from an IP standpoint.
I tend to agree.
AI companies may soon be able to ‘forum shop’ by choosing the most tech-friendly regulatory environments in which to ‘lawfully’ train their AI before rolling out the product internationally – even in jurisdictions where no such usage rights exist.
The more any legislation restricts access to training data for AI machines, the more likely it is to deter developers.
But governments must play a balancing act.
That’s because the more a government opens up AI training possibilities, the more wrath it may feel from the creative industries. The furious reaction from rights owners over the UK’s proposed TDM laws goes some way to showing the strength of feeling.
It’s an exciting time for sure, but lawmakers must be careful not to alienate the creative industries.
And the risks are not just restricted to training.
There is also a danger, remember, that ChatGPT may not generate entirely new content and could even reproduce all or substantial parts of existing copyright-protected works.
The song about the swimming music therapist didn’t appear to, but it doesn’t mean everything else will be free from potential infringement.
It’s not all doom and gloom, though.
Few would have thought that when ChatGPT first launched it would be used by a senior judge within one year.
But that’s exactly what has happened.
Lord Justice Colin Birss, an IP specialist judge at the England and Wales Court of Appeal, revealed at a conference in September that he had used ChatGPT to help him craft part of a judgment.
Birss said he asked the tool to provide a summary of an area of law he already knew, and that he felt the answer was acceptable so used it in his written judgment.
ChatGPT was, he said, “jolly useful”.
Of course, there was plenty of overreaction in some quarters, but it is precisely because ChatGPT is not yet a reliable source that Birss asked it for a summary of a law he already knew.
The fact ChatGPT has gone from an error-strewn, false judgment-citing chatbot to something sophisticated enough for a senior judge to use – all in the space of a year – shows how things could change in the coming months.
Of course concerns remain, but as long as governments are sensible in their lawmaking and ChatGPT’s users don’t get overexcited then the tool should stay on the right side of IP law.
If, however, neither of those things happen then the year ahead could prove very interesting.
In the meantime, there are always those songs about swimming music therapists.