Global AI Regulatory Update - March 2025

In this edition of our global AI bulletin, we will be looking at:

• Global – Global coalition on AI adoption principles in the telecoms industry; an international AI Safety Report; and a Report on AI governance for central banks.
• Asia – Singapore: AI Risk Management Paper; and Hong Kong: New guidelines on AI language models for financial firms.
• Europe – Second draft of General-Purpose AI Code of Practice.
• Middle East – UK and Qatar launch project to boost AI collaboration.
• UK – A report on AI in financial services; and the AI Opportunities Action Plan.
• US – Part two of copyright and AI report published; White House announces development of AI plan; revocation of Biden’s AI Executive Order; new controls on advanced AI technology; release of the Bipartisan AI Task Force report; and an advisory on AI use in derivatives markets.

Global

International AI Safety Report published

On January 29, 2025, the UK’s Department for Science, Innovation and Technology (DSIT) published the first independent International AI Safety Report. The report sets out the first comprehensive, shared scientific understanding of advanced AI systems and their risks. Launched at the AI Safety Summit in November 2023, the report is mandated by over 30 countries including France, China and the US. It brings together insights from 100 independent international experts.

The report highlights the rapid evolution of AI, its potential risks, and the need for reliable design. Key areas for further research include understanding AI capabilities and ensuring reliable behaviour. The report will guide discussions at the AI Action Summit in Paris this year.

Impact: The report aims to provide scientific information that will support informed policymaking, facilitate constructive and evidence-based discussion about the uncertainty of general-purpose AI and its outcomes, and contribute to an internationally shared scientific understanding of advanced AI safety. Businesses should prioritize AI safety, invest in understanding AI systems, and collaborate internationally to mitigate risks.

Report on AI governance for central banks published

On January 29, 2025, the Bank for International Settlements (BIS) published a report on AI adoption in central banks. It highlights the opportunities and risks of AI implementation in critical functions, including data security, confidentiality, and reputational risks. The report recommends using existing risk management models, like the three lines of defence, and adapting them for AI.

The report is a collaborative effort by central banks in the Americas, co-led by representatives from the Bank of Mexico and the BIS.

Impact: The report recommends various practical actions for central banks to adopt AI effectively, including to establish an interdisciplinary AI committee, define principles for responsible AI use, establish an AI framework and update existing guidance, and maintain an AI tools inventory.

Global coalition on AI adoption principles in the telecoms industry

On January 16, 2025, the Department for Science, Innovation & Technology published a notice stating that, through the Global Coalition on Telecommunications, the UK, Australia, Canada, Japan and the US are “cooperating to support innovation in telecommunications technologies in a way that will enable security, growth and societal benefits in all our jurisdictions”. This involves adoption of AI, and the notice sets out principles for the responsible use of AI in telecoms operations.

Impact: The key message of the notice is that the coalition wants to improve network performance, security, and customer experiences by fully integrating AI into future networks. They are also focused on managing risks like security and privacy, working with regulatory bodies, and following international guidelines. Industry efforts and principles like transparency, explainability, and human oversight are key to making sure AI is used responsibly.

Asia

Singapore: AI Risk Management paper published

On December 5, 2024, the Monetary Authority of Singapore (MAS) published an information paper on AI Risk Management. This follows a thematic review of banks’ AI (including generative AI) model risk management practices in mid-2024.

The paper offers guidelines for financial institutions (FIs) to manage AI risks effectively. The recommendations are categorized into governance and oversight, risk management, and development and deployment of AI.

Impact: The MAS information paper provides recommendations for FIs. These include establishing cross-functional oversight forums, updating AI governance policies and procedures, and developing clear AI use statements and principles. Additionally, FIs should identify AI usage and associated risks, maintain inventories of AI applications, and assess the materiality of these risks. It is also important for FIs to ensure data quality and appropriateness, justify and evaluate AI models, focus on explainability and fairness, and conduct independent validations and continuous monitoring.

Hong Kong: New guidelines on AI language models for financial firms

On November 12, 2024, the Securities and Futures Commission (SFC) issued guidelines on the use of generative AI language models by licensed corporations (LCs). These models can enhance efficiency in client interactions and internal processes. However, they pose risks such as inaccuracies, biases, and cybersecurity threats.

The circular emphasizes the need for senior management oversight, model risk management, and robust cybersecurity measures.

Impact: The SFC recommends that LCs implement robust risk management measures. These include model validation, ongoing monitoring, and human oversight for high-risk use cases. LCs are required to notify the SFC of significant changes related to AI language model usage. This ensures compliance and effective governance throughout the AI lifecycle.

Europe

EU: Second draft of General-Purpose AI Code of Practice published

On November 14, 2024, the EU Commission published the first draft of a General-Purpose AI Code of Practice under the AI Act (Act), providing guidance for providers of general-purpose AI models, including those with high impact capabilities that could pose systemic risks. The second draft, reflecting stakeholder feedback, was published on December 19, 2024, with a third draft expected in March 2025.

The Code, prepared by independent experts, is based on principles such as alignment with EU values, proportionality to risks, future-proofing, and support for the AI safety ecosystem. It covers transparency, copyright compliance, and risk assessment for AI models. 

Impact: The Act requires the final version to be published by May 1, 2025, in preparation for the obligations on providers of general-purpose AI models coming into force on August 1, 2025. Compliance with the final-form Code (due by May 1, 2025), will be crucial for all in-scope providers, as it will offer a presumption of conformity with the Act's provisions until formal standards are established. Eversheds Sutherland has introduced an e-learning module to ensure compliance with Article 4 of the Act, which mandates AI literacy for all employees and operators of AI systems from February 2025. This training aims to equip employees with the necessary knowledge to handle AI systems responsibly and ethically.

Middle East

UK-Qatar: UK and Qatar launch project to boost AI collaboration

On December 5, 2024, the UK and Qatar launched a joint AI research commission. Led by Queen Mary University of London and Hamad bin Khalifa University, the project aims to enhance AI collaboration. It will explore ecosystem development, policy, regulation, security, and international engagement.

Funded by the UK government’s Gulf Strategy Fund, the project coincides with the Amir of Qatar’s State Visit to the UK.

Impact: The collaboration seeks to leverage AI for public benefit, addressing emerging risks and fostering innovation. It emphasizes the importance of building strong systems of ethics and governance in AI. It also highlights the importance of international engagement in AI. Businesses should prioritize ethical considerations and responsible innovation in their AI strategies. They should also consider expanding their global partnerships and exploring new markets.

UK

UK: Government launches AI Opportunities Action Plan

On January 13, 2025, the Prime Minister unveiled the UK’s AI Opportunities Action Plan (Plan), together with the government’s response to the Plan. The Plan has three main pillars.

• The first pillar focuses on building secure AI infrastructure, including a new supercomputing facility and AI growth zones. It also aims to unlock public and private data sets for AI training, train and attract AI talent, and ensure safe AI development with expanded regulation.

• The second pillar involves integrating AI into public services, using a “scan > pilot > scale” approach for AI projects, and promoting public-private collaboration in AI adoption.
• The third pillar aims to position the UK as a leader in AI development. It includes establishing the “UK Sovereign AI” for public-private collaboration in frontier AI.

Impact: This is a wide-ranging and ambitious Plan to put AI at the heart of public and private service delivery and to ensure that the UK is a world leader in AI. We wait for more details of how these aspirations will be delivered on, including draft legislation relating to the most powerful AI models. The government says that it plans to consult on proposed legislation to regulate development of the most powerful AI models, but no further detail is given.

UK: Regulators publish joint report on AI in financial services

On November 21, 2024, the Bank of England and Financial Conduct Authority released their third survey findings on AI in financial services. The survey offers continuous insights and analysis on AI usage among regulated firms and financial market infrastructures. This enables regulators to stay informed about the capabilities, development, deployment, and application of AI in financial services.

Key findings of the report include:

• AI is already being used by 75% of respondents
• Another 10% plan to use it over the next three years
• Insurance and international banks have the highest AI usage rates
• Many firms plan to increase AI use for customer support and compliance
• Many firms report only partial understanding of AI due to third-party models
• High regulatory burdens, especially around data protection, pose challenges for stakeholders

Impact: The survey's findings impact financial institutions (FIs) by allowing them to see how their AI usage stacks up against industry standards, which helps pinpoint areas for improvement and innovation. The insights into AI applications are useful for managing risks tied to automated decision-making and third-party dependencies. The survey also guides FIs on where to invest in AI technologies, helping them make smarter strategic decisions.

US

US: Part two of copyright and AI report published 

On January 29, 2025, the US Copyright Office published the second report of a three-part series focusing on copyright and AI. The series analyzes copyright law and policy issues raised by AI. It follows a Notice of Inquiry published in August 2024 seeking comments on copyright and AI systems. Part one of the report looked at digital replicas and was published in July 2024. Part three is expected later this year.

Part two addresses the copyrightability of outputs created using generative AI. It highlights the need for human control over expressive elements in AI outputs. The report emphasizes that copyright protection requires human authorship. AI-generated content without human involvement is not copyrightable. However, works incorporating AI as a tool, where human creativity is evident, can be protected.

Impact: The report concludes that current laws sufficiently address copyrightability and AI issues, making legislative changes unnecessary. Businesses should ensure significant human input in AI-generated works, stay updated on legal developments and court decisions, and understand that prompts alone are insufficient for copyright protection.

US: White House announces development of AI plan

On January 23, 2025, the White House published an Executive Order (EO) titled ‘Removing barriers to American leadership in AI.’ This EO replaced  the Biden Administration’s AI Executive Order titled ‘Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence’, which had been revoked on January 20, 2025 (see our entry below). The new EO mandates the creation of an AI Action Plan within 180 days, involving key advisors and agencies to enhance economic competitiveness and national security.

President Trump's Executive Order (Trump EO) marks a big shift in AI policy from the previous administration. Among other goals, the Biden EO aimed to ensure AI safety, promote innovation, support workers, and protect privacy. The Trump EO, however, revokes these directives, focusing instead on maintaining and enhancing America's global AI dominance.

Impact: This EO is set to transform AI development in the US by removing innovation barriers and ensuring AI systems are “free from ideological bias”. It calls for a review of current policies to align with the new directive, aiming to boost economic competitiveness and national security. Additionally, the National Science Foundation has issued a Request for Information (RFI) to gather public comments on priority policies for the AI Action Plan. The RFI seeks input from a range of stakeholders on a wide range of topics, such as data centers, cybersecurity, and export controls, and will close on March 15, 2025. The AI Action Plan is being directed by key figures like David O. Sacks, and Michael Kratsios. Both have close ties to the tech industry, which could present potential conflicts of interest.

US: Trump revokes Biden’s AI Executive Order

On January 20, 2025, President Trump revoked several Executive Orders from the Biden administration, among them Executive Order 14110, titled ‘Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence’ (Biden EO). Its fact-sheet explains that revocation was needed to remove unnecessary burdens on companies working with AI and to encourage the development of AI systems “free from ideological bias or engineered social agendas”.

Impact: The impact of revoking the Biden EO remains to be seen, but we expect significant effects, especially on private sector entities involved in AI development. Reporting requirements for advanced AI models and standards for AI safety, previously mandated by the Biden EO, may be rescinded or revised. The AI Safety Institute, established under Biden, may face cuts or closure.

US: New controls on advanced AI technology

On January 13, 2025, the Biden-Harris administration announced new controls on advanced AI technology to protect national security. The framework includes controls on advanced computing chips and closed AI model weights. It introduces license exceptions for allies and partners, ensuring innovation is not stifled. However, it is unclear if these controls will survive the new administration.

The regulation supports the government’s strategy to cultivate a secure technology ecosystem for the responsible use and diffusion of AI.

Impact: The rule aims to prevent AI misuse by adversaries while promoting responsible AI diffusion globally. It is designed to balance security with technological leadership, ensuring AI benefits are shared with trusted entities. Businesses should stay informed on evolving AI regulations to mitigate risks and capitalize on opportunities.

US: Bipartisan AI Task Force report released

On December 17, 2024, the House of Representatives AI Task Force published a report with input from relevant committees. It outlines principles and recommendations for responsible AI innovation, emphasizing the importance of maintaining US leadership in AI while having appropriate safeguards. With the new Republican leadership, the AI Task Force is likely to align any legislative efforts with the White House’s AI Action Plan.

The report emphasizes the importance of maintaining America’s leadership in AI while ensuring appropriate safeguards. The report aims to inform Congress and the public about AI's benefits and risks, serving as a foundation for future legislation.

Impact: The report aims to protect consumers, foster innovation, and secure US leadership in AI technology, serving as a foundation for future legislative actions to address AI-related challenges and opportunities. Financial regulators are expected to foster responsible AI adoption in financial services, increase AI expertise, maintain consumer and investor protections, consider regulatory "sandboxes" for AI experimentation, support a principles-based regulatory approach, and ensure regulations do not impede small firms from adopting AI.

US: Regulator publishes advisory on AI use in derivatives markets

On December 5, 2024, the Commodity Futures Trading Commission (CFTC) issued an advisory on AI use in regulated markets. The CFTC states that entities must comply with the Commodity Exchange Act and CFTC regulations. The advisory highlights potential AI use cases in areas such as order processing, trade matching, and market surveillance.

The CFTC will continue to monitor AI developments and their potential benefits and risks. This is informed by public comments from a January 2024 request. The advisory is part of the CFTC’s ongoing oversight and dialogue with regulated entities.

Impact: It is unclear if the CFTC will survive the new administration's reorganization plans. Press reports indicate that President Trump wants to merge the CFTC with the SEC, which would require legislative action. Assuming the CFTC survives this reorganization, any CFTC advisory will also be revised in light of the White House AI Action Plan. Financial institutions should stay informed about these developments and be prepared to adapt to any changes in regulatory oversight.

Co-authored by Jon Botham and Kirath Bharya (Knowledge).

Further reading

• Trump Administration Executive Orders & Legal Developments
• Defining AI systems | Illuminating your EU AI Act compliance needs | Eversheds Sutherland
• AI Opportunities – UK AI Action Plan Unveiled
• UK Government consults on copyright and AI: Have your say!
• U.S. and China tighten respective export restrictions on advanced technology and critical minerals | Eversheds Sutherland
• AI and ESG: Emerging forces shaping growth strategies and M&A | Telescope | Eversheds Sutherland
• Navigating AI and data protection | Illuminating your EU AI Act compliance needs | Eversheds Sutherland
• Geopolitical dynamics and AI: The tug-of-war over development and regulation around the globe | Eversheds Sutherland
• Global Legal Tech Series 2025 | Eversheds Sutherland
• Understanding AI Literacy: What your organization needs to know | Illuminating your EU AI Act compliance needs | Eversheds Sutherland
• Evolving trade rules: Using AI to integrate ESG goals into the global trading system | Telescope | Eversheds Sutherland
• What Trump 2.0 could mean for AI policy and the global technology sector | Tech Talks - Episode 4 | Eversheds Sutherland

---