Artists Slam UK Plan To Let AI Train on Copyrighted Works by Default

Key Takeaways

• The U.K. government has outlined proposals to exempt AI training from copyright protections.
• Under the proposed system, AI training would be allowed by default unless rights holders expressly opt-out.
• Critics argue that requiring content creators to opt out of AI training may fail to prevent the unapproved use of copyrighted works.

Amid ongoing disputes over the use of copyrighted material to train AI, the U.K. government has proposed amending intellectual property law in the country to clarify the rules.

Under the proposal, data mining for AI training would be exempt from copyright obligations unless rights holders specify otherwise.

However, after a six-week consultation, a clear rift has emerged between the government and creative industries.

U.K. Explores Options for Copyright Law

Citing the current uncertainty surrounding AI copyright law, in December the government outlined three different policy options:

Option one would require AI developers to gain an express license from copyright owners to use their content.

The second would provide a broad “data mining exemption” that permits AI training without restrictions.

The third option, preferred by the government, would let AI developers train models on copyrighted materials unless rights holders explicitly opt-out.

This “would mean that AI developers are able to train on large volumes of web-based material without risk of infringement,” the government stated.

“Importantly, right holders are also able to control the use of their works using effective and accessible technologies and seek payment through licensing agreements,” the notice read.

In theory, the government’s copyright proposal should empower creators to withhold their content from AI training libraries.

However, after a six-week consultation period that ended on Feb. 25, it is clear that many stakeholders oppose the plans.

Creatives Criticize U.K. Government’s Approach

Criticism of the government’s proposal has mostly emerged from the creative sector.

Musician Kate Bush said the government’s “willingness to agree to these copyright changes shows how much our work is undervalued.”

Similarly, Elton John and Paul McCartney have joined calls for stronger copyright protections against data mining for AI.

Many object to the proposals on the grounds that without a requirement for explicit content, it will be almost impossible to prevent content from being hoovered up by web scrapers.

“There are currently no viable rights reservation solutions that would give artists meaningful control over the use of their work,” the Design and Artist’s Copyright Association (DACS) pointed out in its response to the government consultation.

“Especially when art and images are readily distributed by third parties” DACS added.

Opt-Out Unworkable, Critics Claim

DACS concerns were echoed by Vermillio, a company that makes content provenance tools.

Responding to the government’s consultation, Vermillio was critical of the notion that right holders could simply opt out of their material being used by AI developers:

“Opt-out processes inevitably lead to content being used for AI training whose right holders or creators don’t want them to be used.”

“Intellectual property rights are not binary in complex markets,” it added. “they consist of multiple, divisible rights that traditionally must each be negotiated through commercial negotiations.”

Calls for Greater Transparency

One concern that arose in many responses to the government’s consultation was a lack of transparency.

Regardless of whether the U.K. pursues an opt-in or opt-out approach, stakeholders argued that copyright holders should be able to ascertain when their content has been used to prevent abuse by AI developers.

Transparency is important not just for IP law but for data protection, too.

In its response to the consultation, the Information Commissioner’s Office (ICO) argued that greater transparency from AI developers is “urgently needed.”

As the regulator pointed out, some copyrighted materials are also legally protected personal information.

Should the government pursue the opt-out approach, “it will be important to provide clarity that this will not in and of itself constitute a determination of the lawful basis for any personal data processing,” the ICO stressed.