Widespread concern about the usage of artistic works to coach synthetic intelligence (AI) programs has prompted the UK authorities to start exploring how the nation’s copyright guidelines will be modified to fulfill the complicated, usually conflicting calls for of each the artistic and tech sectors.
The choices being assessed embrace protecting copyright and associated legal guidelines stay as they’re; strengthening copyright to require licenses in all instances; implementing a broad information mining exemption for AI firms; or making a extra restricted information mining exemption that enables copyright holders to order their rights, underpinned by measures to advertise and assist larger transparency from builders.
Nonetheless, given structural imbalances inside present copyright markets – which favour big firms over particular person creators – it’s unclear to what extent the AI-related reforms to the UK mental property guidelines being thought-about will assist artistic staff themselves.
Creators vs AI builders
Questions round the usage of artistic works to coach AI programs have grow to be one of the crucial intense areas of debate because the creation of generative AI (GenAI) and huge language fashions (LLMs) with the discharge of OpenAI’s ChatGPT in November 2022.
Specifically, the controversy has targeted on what it means for present copyright protections and the livelihoods of creators, who’ve expressed concern over the unauthorised use of their works to coach AI fashions.
Other than an absence of transparency from AI firms concerning the information included of their coaching corpuses, creatives have variously complained concerning the absence of enforceable mechanisms to guard their copyrighted works throughout the context of scraping at scale, in addition to the impacts of AI on artistic job markets and competitors.
For AI firms, alternatively, entry to huge quantities of high-quality information is of paramount significance, significantly relating to the event of LLMs comparable to Claude, ChatGPT or Gemini.
A submission to the US Copyright Workplace on 30 October 2023 by Amazon and Google-backed LLM developer Anthropic is indicative of how these corporations view their use of copyrighted materials, and the way integral it’s for creating generative AI fashions.
“To the extent copyrighted works are utilized in coaching information, it’s for evaluation (of statistical relationships between phrases and ideas) that’s unrelated to any expressive goal of the work,” it stated. “This form of transformative use has been recognised as lawful prior to now and will proceed to be thought-about lawful on this case.”
It added that utilizing copyrighted works to coach its Claude mannequin would depend as “honest use”, as a result of “it doesn’t stop the sale of the unique works, and, even the place business, remains to be sufficiently transformative”.
As a part of a separate authorized case introduced in opposition to Anthropic by main music publishers in November 2023, the agency took the argument additional, claiming “it could not be potential to amass adequate content material to coach an LLM like Claude in arm’s-length licensing transactions, at any worth”.
It added that Anthropic is just not alone in utilizing information “broadly assembled from the publicly obtainable web”, and that “in apply, there isn’t a different solution to amass a coaching corpus with the dimensions and variety crucial to coach a posh LLM with a broad understanding of human language and the world typically”.
If licences had been required to coach LLMs on copyrighted content material, in the present day’s general-purpose AI instruments merely couldn’t exist Anthropic
“Any inclusion of plaintiffs’ tune lyrics – or different content material mirrored in these datasets – would merely be a byproduct of the one viable method to fixing that technical problem,” it stated.
It additional claimed that the dimensions of the datasets required to coach LLMs is just too massive to for an efficient licensing regime to function: “One couldn’t enter licensing transactions with sufficient rights homeowners to cowl the billions of texts essential to yield the trillions of tokens that general-purpose LLMs require for correct coaching. If licences had been required to coach LLMs on copyrighted content material, in the present day’s general-purpose AI instruments merely couldn’t exist.”
Whereas the submission and the court docket case are particular to the US context, the appliance of “honest use” exemptions to copyright is just not dissimilar in UK. Beneath present UK copyright legal guidelines, unique works are routinely protected upon their creation, giving the creators unique rights to repeat, distribute, carry out or adapt their work.
There are, nevertheless, restricted exemptions that enable the “honest dealing” of copyrighted materials for the needs of, for instance, analysis, criticism, evaluation and reporting. An extra exemption was added in 2014, permitting textual content and information mining for purely non-commercial analysis functions.
Because it stands, except one among these exemptions applies, AI firms would due to this fact have to acquire permission from copyright holders to make use of these works of their mannequin’s coaching information.
UK authorities session backlash
In response to a earlier UK authorities session on the matter, which closed in February 2025, “the appliance of UK copyright regulation to the coaching of AI fashions is disputed”.
It stated that whereas rights holders are discovering it tough to manage the usage of their works in coaching AI fashions, and are in search of to be remunerated for its use, AI builders are equally discovering it tough to navigate copyright regulation within the UK. It famous “this authorized uncertainty is undermining funding in and adoption of AI know-how”.
In an try to resolve the dispute, the UK authorities proposed a brand new coverage in late 2024 that may enable AI firms to coach their fashions on copyrighted works except rights holders explicitly opted out. Because of this, relatively than requiring AI firms to hunt permission from rights holders for the usage of their work, the burden could be positioned on the creators themselves to actively object.
The opt-out proposal provoked vital backlash from creatives, who seen it as too conciliatory to the slim pursuits of tech firms. Out of the greater than 10,000 people who responded to the federal government’s session on these measures, simply 3% backed it’s opt-out proposal, whereas 95% referred to as for both referred to as for copyright to be strengthened, a requirement for licensing in all instances, or no change to present copyright regulation.
Others cited points across the practicality of such proposals, noting that within the context of the present digital panorama – the place copyrighted content material is scraped at scale and included in coaching datasets, usually with out attribution – it could be unimaginable for somebody to know when their work has been used, not to mention decide out.
Within the wake of this widespread opposition, the UK authorities has since dedicated to exploring a licence-first system that may require AI firms to hunt express permission from creatives and supply them with compensation.
Balancing pursuits?
A 12 months later, in December 2025, know-how secretary Liz Kendall advised Parliament there was “no clear consensus” on the AI-copyright problem, saying that the federal government would “take the time to get this proper” whereas promising to make coverage proposals by 18 March 2026.
“Our method to copyright and AI should assist prosperity for all UK residents, and drive innovation and development for sectors throughout the economic system, together with the artistic industries,” she stated. “This implies protecting the UK on the chopping fringe of science and know-how so UK residents can profit from main breakthroughs, transformative innovation and larger prosperity. It additionally means persevering with to assist our artistic industries, which make an enormous financial contribution, form our nationwide identification and provides us a novel place on the world stage.”
Whereas authorities rhetoric on AI and copyright has revolved round the necessity to assist each the UK’s artistic and tech sectors, there’s a sense that – to this point at the very least – it’s prioritising the latter in its ambition to make the nation a tech superpower.
Beeban Kidron, a crossbench peer and former movie director, for instance, has beforehand described the usage of copyrighted materials by AI firms as “state-sanctioned theft”, claiming ministers could be “knowingly throwing UK designers, artists, authors, musicians, media and nascent AI firms underneath the bus” in the event that they don’t transfer to guard their output from being harvested by AI corporations.
Owen Meredith, chief government of the New Media Affiliation, has additionally beforehand urged the UK authorities to rule out any new copyright exception. “This can ship a transparent message to AI builders that they have to enter into licensing agreements with the UK’s media and inventive copyright homeowners, unlocking funding and strengthening the marketplace for the high-quality content material that’s the most beneficial ingredient in producing secure, reliable AI fashions,” he stated.
Ed Newton-Rex, a outstanding commentator on AI and mental property, has additionally criticised the stability of UK authorities’s method, noting that whereas the federal government described its session proposals on the time as a “win-win … that is very removed from the reality. It might be an enormous coup for AI firms, and probably the most damaging laws for the artistic industries in many years”.
He added {that a} broad copyright exception that enables unlicensed coaching on copyrighted “would hand the life’s work of the UK’s creators to AI firms, letting them use it to construct extremely scalable rivals to these creators with impunity”.
[A broad copyright exception] would hand the life’s work of the UK’s creators to AI firms, letting them use it to construct extremely scalable rivals to these creators with impunity Ed Newton-Rex, AI and mental property commentator
AI firms, in fact, disagree. In its October 2023 submission to the US Copyright Workplace, Anthropic argued that requiring licences could be inappropriate, as it could lock up entry to the overwhelming majority of works and profit “solely probably the most extremely resourced entities” which are capable of pay their manner into compliance
“Requiring a licence for non-expressive use of copyrighted works to coach LLMs successfully means impeding use of concepts, information and different non-copyrightable materials,” it stated. “Even assuming that features of the dataset could present larger ‘weight’ to a specific output than others, the mannequin is greater than the sum of its components. Thus, it will likely be tough to set a royalty charge that’s significant to particular person creators with out making it uneconomical to develop generative AI fashions within the first place.”
Others from the tech sector have additionally argued that diverging from different jurisdictions too enormously – for instance, by implementing a UK-specific licensing association most well-liked by the artistic sector, or requiring corporations to reveal detailed information inputs – would merely imply AI firms keep away from deploying within the UK.
Commerce affiliation TechUK, for instance, argued that within the context of AI-copyright associated amendments to the federal government Knowledge Use and Entry Invoice – which might have pressured builders to publish their coaching corpuses however which had been in the end not included within the closing Act of Parliament – departing an excessive amount of from present UK and worldwide frameworks would dangers firms being “discouraged from working, coaching and deploying AI merchandise and fashions within the UK”.
This was additionally recognised by the federal government in its session, which famous requiring licenses in all instances “is extremely more likely to make the UK considerably much less aggressive in comparison with different jurisdictions – such because the EU and US – which do not need such restrictive legal guidelines. This may make the UK a much less engaging location for AI improvement, decreasing funding within the sector. In doing so, it could not really enhance the extent of licensing undertaken by AI corporations.”
It added that fashions educated in different jurisdictions which don’t meet any new UK requirements could also be tough to limit from the UK market, and dangers among the most succesful AI fashions not being made obtainable within the UK: “This may considerably restrict innovation, client alternative and wider advantages of AI adoptions throughout the UK economic system.”
The technical caveats of copyright regulation
Beneath UK copyright regulation, it needs to be famous that creating “transient copies” of works is allowed if sure circumstances are met. This consists of if it’s not a everlasting copy and serves a short, ancillary goal; if it’s a crucial step in a technological course of; if its solely purpose is enabling lawful use or community transmission; and the copy itself doesn’t maintain separate business worth.
When AI mannequin coaching processes – which regularly, however not at all times, retain solely a really small portion of every coaching merchandise – this means it could be technically flawed to say a copyright infringement has taken place, as Anthropic has argued within the context of the US.
Nonetheless, this doesn’t imply {that a} mannequin would by no means infringe copyright, as additionally it is technically potential for many fashions to “memorise” copyrighted works, turning a transitory copy right into a everlasting, infringing one.
Though the specificities of whether or not a specific mannequin or AI-generated output infringes this present copyright regime will probably be hashed out in particular person court docket instances, some have argued that in search of copyright to resolve the stress between creatives and AI firms is a non-starter.
In essence, their argument is that increasing copyright may be very unlikely to guard the roles or incomes of already underpaid creatives, who’ve themselves been exploited by leisure behemoths wielding copyright legal guidelines in opposition to them for many years.
Of their Could 2024 e book, Who owns this sentence? A historical past of copyright and wrongs, authors Alexandre Montagu and David Bellos equally argue that copyright protections – which had been initially supposed to guard the livelihoods of particular person creators – have since been transferred to massive firms as a substitute, which use them to extract a type of “lease” from customers globally, whereas additionally locking the workers who helped contribute to the creation of the IP out from possession and the resultant advantages.
It follows, then, that there’s little cause to consider these similar firms will now deal with their artistic staff extra pretty in the event that they obtain compensation as a copyright holder from AI firms.
To change this dynamic, Doctorow and others argue it could require altering the very construction of creatives markets in order that the advantages accrue to creatives, relatively than massive firms that primarily run “tollbooths” to facilitate and management entry to artistic’s work, which in flip permits them to extract disproportionally excessive earnings for themselves.
Writing for the Digital Frontier Basis (EFF) in February 2025, Tori Noble argued that “increasing copyright is not going to mitigate” the hurt to artistic staff, and that “what neither Huge Tech nor Huge Media will say is that stronger antitrust guidelines and enforcement could be a a lot better resolution”.
She added that trying past copyright can future-proof protections, together with stronger environmental protections, complete privateness legal guidelines, employee protections and media literacy, including: “[This will] create an ecosystem the place we may have defences in opposition to any new know-how that may trigger hurt in these areas, not simply generative AI. Increasing copyright, alternatively, threatens socially helpful makes use of of AI – for instance, to conduct scientific analysis and generate new artistic expression – with out meaningfully addressing the harms.”
Collective copyright and labour regulation
Because it at present stands, UK authorities appears to be on track to introduce a brand new licensing regime for AI firms’ use of copyrighted supplies. Observers have stated this would wish to incorporate mechanisms that enable creators to determine when and the way their works are used, in addition to to object or search compensation as they see match.
Nonetheless, given the clear tensions that exist already between particular person and company copyright holders, even a licensing regime might nonetheless disproportionally profit the latter. It might additionally disproportionally profit massive AI builders, because the pool of actors with the power to pay for sufficient copyright licenses to successfully practice a mannequin is vanishingly small.
Using AI in artistic endeavours throws up additional points round labour and competitors: even when creators acquired compensation for the usage of their copyrighted materials, AI’s whole improvement is underpinned by a neoliberal logic of austerity. Because of this, within the present political-economic context, these with the decision-making energy to deploy AI largely accomplish that as a result of it permits them to chop labour prices – the largest overhead for any capitalist enterprise.
Within the present political-economic context, these with the decision-making energy to deploy AI largely accomplish that as a result of it permits them to chop labour prices – the largest overhead for any capitalist enterprise
In November 2024, information from the Harvard Enterprise Overview confirmed the influence that generative AI fashions had been already having on labour markets, which highlights how creatives will primarily find yourself competing with the very fashions that ingest their information. Particularly, it highlighted how the introduction of ChatGPT decreased writing and coding jobs by 30% and 20% respectively, whereas AI picture turbines equally decreased picture creation jobs by 17%.
Given the sheer scale at which fashions ingest information, it’s not onerous to see how creatives – even with a licensing regime in place – could possibly be undermined by bosses who would relatively pay for a comparatively low-cost company licence to an AI mannequin, relatively than the comparatively costly labour of human beings.
Within the tech sector itself, corporations globally have been busy chopping their workforces as they appear to extend spending on and funding in AI instruments. In October 2025, Amazon laid off 14,000 staff, a choice that was particularly prompted and enabled by the agency’s AI investments.
Whereas many argue that the appearance of AI is inevitable, its impacts are actually not. In November 2023, for instance, the Autonomy assume tank within the UK argued that whereas automating jobs with LLMs might result in vital reductions in working time with no lack of pay or productiveness, realising the advantages of AI-driven productiveness positive factors on this manner would require concerted political motion.
The assume tank added this was as a result of it’s clear that productiveness positive factors usually are not at all times shared evenly between employers and staff, and rely on “geographic, demographics, financial cycle and different intrinsic job market components” comparable to staff’ entry to collective bargaining.
To ship optimistic AI-led modifications for staff and never simply employers, Autonomy advisable establishing “automation hubs”, underpinned by commerce union and business agreements, to spice up the adoption of LLMs in methods which are equitable.
Within the context of the artistic industries and copyright, the same scenario has already taken place with the 2023 Hollywood writers’ strike, whose collective sector-wide motion ended with an settlement from studios that AI can’t be used to write down or rewrite scripts, and which gave them the power to ban the usage of their writing in mannequin coaching.
As a substitute of replying purely on copyright regulation – which traditionally has been wielded in opposition to particular person creatives by leisure and media firms – the reply could also be present in trying to construct up collective copyright mechanisms and bettering the underlying labour protections for artistic staff to cease them being ripped off by firms, with or with out the assistance of AI.
