The Communications and Digital Committee of the House of Lords of the UK recently published its fourth report, Large language models and generative AI. This report serves as a milestone in the Committee's series of inquiries into the intersection of generative AI and copyright law.

This report is part of the UK government's long-term efforts to address the widespread concerns of the creative and AI industries, incorporating feedback and evidence from numerous stakeholders.

While the report primarily focuses on the UK framework, it also offers valuable insights for Canadian policymakers and relevant practitioners. Several of the issues explored are currently being studied by the Canadian government, including text and data mining (TDM), transparency regarding works used for AI model training, and the protection of creator economic rights in the era of generative AI.

Recommendations

The Committee proposed several measures to the UK government in the report, which are summarized as follows:

1. Reject amendments to the UK Copyright, Designs and Patents Act 1988 and refuse to introduce a commercial text and data mining exception.

During previous public consultations, the UK government drafted several response strategies to TDM activities. One proposal was to establish a broad TDM exception for commercial purposes, regardless of whether a rights reservation mechanism (i.e., opt-out) was in place. In the report, the Committee recommended that the government discard this option.

Unlike Canada, the UK Copyright, Designs and Patents Act 1988 stipulates that TDM copies made for non-commercial research purposes may be covered by an exception, provided the content is lawfully accessed and properly attributed where possible. Additionally, the Act provides a general exception for temporary copies generated during technical operations, a rule similar to Section 30.71 of the Canadian Copyright Act.

The Committee recommended that the UK government should not introduce any legislative changes that weaken the willingness to authorize the use of copyrighted works for AI training, including any exceptions for commercial TDM. The Committee believes that the tech industry’s push for such an exception is fundamentally an attempt to reduce its own litigation risks by weakening copyright protection, rather than a neutral consideration of legal clarification. In the Committee's view, the large-scale reproduction and processing of copyrighted works to train AI models constitute reproduction and should follow general copyright principles and be judged within the existing framework of exceptions.

The Committee urged the UK government to focus on improving licensing mechanisms, information disclosure rules, and enforcement within the existing legal system. It also recommended that the government issue a public statement as soon as possible, explicitly requiring commercial AI developers operating in the UK to obtain appropriate authorization according to the law when using copyrighted works to train generative AI models.

2. Address protection gaps regarding identity, style, and digital replicas (including the use of "style-alikes").

The Committee noted that the current UK law lacks a comprehensive personality rights system and does not provide specific protection for digital likenesses. If a product mimics the unique style, voice, or personal likeness of a creator or performer without copying the original content, rights holders struggle to defend their rights under the law. In response, the Committee recommended adding new legislative protections to prohibit the unauthorized creation of digital replicas and the imitation of another’s creative style, allowing creators and performers to legally control the commercial use of their likeness while reasonably safeguarding freedom of speech and other fair uses.

3. Make the disclosure of AI training data information a statutory obligation.

The Committee stated that general disclosure by AI developers regarding the materials used for model training fails to satisfy the demands of rights holders; therefore, more granular information disclosure requirements must be implemented.

The Committee is aware of potential disagreements among AI companies regarding the scope of disclosure and therefore recommended that the UK government take the lead in organizing consultations between AI companies and copyright holder representatives to develop reasonable and operational solutions. Simultaneously, a dedicated regulatory body should be designated to set unified statutory information disclosure standards for large AI companies. The design of these rules should aim to avoid scenarios where local enterprises transfer training operations overseas to circumvent compliance requirements, thereby harming the interests of UK creators, innovators, and consumers.

4. Promote the development of technical standards for data governance, source traceability, and content labeling.

The Committee emphasized that effective machine-readable rights reservation mechanisms are key to building a sustainable licensing system. Although existing website-level reservation tools cannot meet the needs of modern AI systems, the Committee noted that various solutions have emerged in the industry, including encrypted metadata, content fingerprinting, and digital watermarking, and recommended that relevant technologies be studied and adopted.

During this transition, referencing Australia’s regulatory approach, the Committee recommended that the UK government publicly state that it will not introduce a new commercial TDM exception based on an "opt-out" rights reservation model. Furthermore, the Committee clarified that clear, prominent labeling of AI-generated content is at the core of the UK’s AI and copyright strategy. This measure can maintain the value of human creative achievements, help consumers identify the source of content, and facilitate fair competition in creative markets that prioritize original works.

5. Foster a fair and inclusive AI copyright licensing market.

Several leading AI companies and creative institutions have recently reached partnership agreements, signaling that a content licensing market for AI business has gradually taken shape. The Committee recommended that the AI licensing system built by the UK should be compatible with multiple licensing models, balancing the needs of copyright holders of different scales with those of AI companies. Collective management organizations can play a significant role in assisting creators with licensing negotiations and revenue collection. This may include exploring the introduction of an inalienable right to equitable remuneration when a rights holder’s works and performances are used as input (and in appropriate cases, output) for AI training, with all such uses subject to mandatory collective management; it also supports creator-first remuneration models, supplemented by appropriate transparency and audit arrangements.

6. Prioritize the R&D and application of autonomous AI models.

The Committee recommended that the UK should not remain reliant on non-transparent AI models developed in the US. As the government advances the development of domestic autonomous AI, it should prioritize the research, development, and promotion of models controlled by the UK that ensure transparency from the design stage, clearly publicizing information related to training data and R&D processes.

Related Developments in Canada

In October 2023, the Canadian government launched a copyright consultation in the era of generative AI, aimed at soliciting views and evidence from stakeholders on key copyright issues triggered by generative AI, subsequently publishing the "What We Heard" report summarizing the feedback received. The Canadian AI Strategy Working Group also published the New Generation AI Strategy Communication Summary in Q1 2026, recommending the formulation of a national AI talent strategy, the establishment of AI assurance labs and governance committees, the conduct of goal-oriented research aligned with public needs, and the strengthening of intellectual property protection alongside structural reforms to support the growth and retention of Canadian AI enterprises.