The UK’s House Of Commons yesterday passed the Data Bill without any amendments addressing issues around copyright and AI, despite attempts by MPs to add a section that would force AI companies to comply with various copyright and transparency obligations.
Which means there won’t be any imminent changes to UK law regarding copyright and AI, despite many people from across the creative industries arguing that urgent measures are needed to stop AI companies hoovering up huge quantities of copyright protected material to train generative AI models without the permission of copyright owners.
The measures proposed by MPs would have imposed a legal obligation on AI companies to say what content they used to train their models, and would also have clarified that AI companies operating in the UK must comply with UK copyright law, which means getting permission from copyright owners
Those measures were first added to the Data Bill when it was being debated in the UK Parliament’s upper chamber, the House Of Lords. But when the legislation moved to the House Of Commons, the government got the Lords’ amendments removed.
In yesterday’s Commons debate opposition MPs tried to get them put back in, but that attempt was ultimately unsuccessful. The Lords may well have one more go at getting the amendments included when they consider the final draft of the bill, but it seems likely the government will get its way on all this.
The government’s main justification for rejecting the amendments is that all the issues around copyright and AI should be addressed in one place through specific legislation informed by its recent consultation on this debate, rather than by rushing through measures to address certain specific issues as part of other legislation.
That approach allows for more time to properly review all the issues and all the proposed solutions - and to also consider related topics beyond copyright, like the role of personality rights in helping performers protect their likeness and voice.
Defending this position yesterday, culture minister Chris Bryant said, “It must surely be better to legislate on this complex subject in the round rather than piecemeal”.
Another reason for delaying this is that the government is still hoping to find solutions to these issues that protect the UK’s creative industries while also providing AI companies the flexibility they say they need, with ministers still seemingly optimistic that that can be achieved.
On one level the government’s approach makes sense, although there is currently no clear timeline for when any bespoke legislation on copyright and AI may be ready to present to Parliament. There is, however, a commitment to report back to MPs on possible solutions within one year of the Data Bill becoming law.
There’ll be four reports in total, one covering technical solutions to allow rightsholders to manage their rights in the context of AI; one looking at how AI companies can best access the data they say they need; one on transparency and one on licensing.
In his speech yesterday, Bryant told MPs that the government will put “binding commitments” into the Data Bill to “assess the impact of any and all” proposed solutions.
The proposal that has got the most attention in recent months is the introduction of a new text and data mining exception with an opt-out, allowing AI companies to make use of copyright protected works without getting rightsholder permission, except where rightsholders have formally opted out.
The government basically said that was its preferred solution even before opening its consultation of copyright owners and AI companies.
That proposal is strongly opposed by the creative industries and, in recent days, ministers have indicated it’s no longer their preferred solution, although it is still on the table.
Last weekend The Guardian reported that technology minister Peter Kyle and his team were now considering various options without favouring one over the other, while The Telegraph yesterday said that Kyle is now most keen on masterminding some sort of licensing scheme that works for everyone.
But in Parliament yesterday Bryant was keen to stress that until the civil servants have ploughed through the 11,500 submissions that were made to the AI consultation, all options remain open.
“The government must keep an open mind, and must take full account of the economic evidence”, he told MPs, adding that the government has “not prejudged the outcome of the consultation”, and is aware that it must “consider and reflect on the best approach for all parties”.
Interestingly, Bryant also reminded MPs that there are more issues to be discussed than the exception-with-opt-out proposal and transparency requirements. While that is true, it was the government’s own consultation that pushed the exception and transparency to the top of the agenda.
Nevertheless, one of those other issues is very relevant to performers in the music industry, which is the ability to protect voice and likeness in the context of AI. Many have argued that so called personality rights are most relevant when it comes to performers stopping people using AI to imitate their voice and likeness, but these rights don’t currently exist in the UK.
“Some issues have hardly been referred to in the public debate on this matter”, Bryant told MPs. That includes, he added, personality rights. “That is quite complicated to legislate for”, he went on, adding that submissions to the consultation are informing a way forward.
Which is why the government continues to argue that all of the different issues around copyright and AI should be properly considered and then dealt with in a joined up way.
There is a logic to that argument, though for some in the creative industries there is a real sense of urgency in this debate. And for them, waiting another year for some reports and then who knows how much longer for legislative proposals - while AI companies continue to scrape large quantities of unlicensed content off the internet - just isn’t good enough.