As we usher in 2025, we look back on a groundbreaking year that redefined the landscape of artificial intelligence. The past 12 months have been nothing short of revolutionary, with AI advancements surpassing even the most optimistic predictions. Generative AI, in particular, has evolved from an experimental technology to a cornerstone of business strategy, with multimodal capabilities pushing the boundaries of what's possible. The integration of AI into core business workflows has become not just a competitive advantage, but a necessity for staying relevant in an increasingly AI-driven market. As we look ahead to the next 12 months, we catch you up on some of the more eye-catching developments of the past month.
European initiatives
EDPB issues opinion on AI and data protection
The European Data Protection Board (EDPB) released Opinion 28/2024, addressing critical data protection aspects in AI model development and deployment. The opinion focuses on three main areas:
- AI model anonymity;
- the use of legitimate interest as a legal basis for processing personal data in AI contexts; and
- the implications of using unlawfully processed data in AI models.
For AI system developers, the opinion sets a high bar for demonstrating model anonymity and emphasises the need for robust technical measures to mitigate re-identification risks. The EDPB also outlines stringent requirements for relying on legitimate interest, including a necessity assessment and implementation of data minimisation techniques. Developers must be particularly cautious about web scraping practices, ensuring selective data collection and avoiding sensitive or intrusive sources. Additionally, the opinion stresses the potential legal consequences of using unlawfully processed data, highlighting the importance of due diligence in data sourcing and model development.
EU General-Purpose AI Code of Practice - second draft published
The European AI Office released the second draft of the General-Purpose AI (GPAI) Code of Practice, designed to guide model providers in complying with the upcoming EU Artificial Intelligence Act. This draft incorporates feedback from nearly 1,000 stakeholders and focuses on key areas such as transparency, copyright compliance, risk assessment and governance. It mandates comprehensive documentation of AI models, including data sources and authorisations, and emphasises internal policies for lawful data usage. Additionally, providers of advanced GPAI models must conduct mandatory risk assessments and implement incident reporting procedures. The Code features a modular structure with objectives and performance indicators for compliance measures, facilitating flexibility while ensuring accountability. The drafting process will continue with a final version expected by April.
EU General Product Safety Regulation comes into effect
The European Union's General Product Safety Regulation (GPSR), which came into force on 13 December 2024, marks a significant update to consumer product safety frameworks, particularly in addressing AI-related risks. As part of the EU's comprehensive approach to AI regulation, the GPSR complements the AI Act by introducing specific requirements for AI-powered consumer products. While retaining the broad definition of "product" from its predecessor, the General Product Safety Directive, the GPSR now explicitly covers AI-enabled consumer items but excludes standalone AI software. The regulation imposes new obligations on economic operators, including manufacturers, to ensure the safety of AI-integrated products before they enter the EU market. Key provisions include enhanced risk assessment procedures, standards for online marketplaces, and improved mechanisms for product recalls and consumer information.
Key EU AI Act deadlines approaching in 2025
The EU AI Act, which entered into force on 1 August 2024, introduces several critical deadlines for AI system providers and deployers in 2025. The first major milestone arrives on 2 February 2025, when prohibitions on unacceptable risk AI practices take effect, including bans on social scoring, biometric categorisation and certain uses of facial recognition technology. By 2 May 2025, the European Commission is expected to have codes of practice ready for AI providers. 2 August 2025 marks another significant date, as rules for general-purpose AI models, governance structures and penalties become applicable. Providers of general-purpose AI models placed on the market before this date will have until 2 August 2027 to ensure compliance.
UK initiatives
ICO releases final response on generative AI and data protection
The UK Information Commissioner's Office (ICO) published its final response to the consultation on generative AI and data protection, which began in January 2024. This response addresses key regulatory uncertainties related to UK GDPR and the Data Protection Act 2018, focusing on areas such as lawful data usage for web scraping, purpose limitation, accuracy of training data, individual rights and controllership. The ICO emphasises the need for improved transparency from generative AI developers regarding training data sources and refines its guidance on lawful bases for using web-scraped data. Additionally, it underscores that data protection principles apply to new technologies and that major AI developers must demonstrate ongoing compliance. The ICO plans to update its guidance to incorporate these positions, aiming to provide regulatory clarity while fostering responsible AI innovation.
Human rights concerns raised over AI Judges
Giving the Blackstone Lecture at Pembroke College, Oxford, the UK’s most senior civil law judge called for the establishment of a human right to have legal cases heard by human judges rather than AI systems. Sir Geoffrey Vos, the Master of the Rolls, who sits regularly on the Court of Appeal, said that allowing machines to make court decisions would be an “existential challenge to our humanity” and that governments should “ensure that the activities within the life cycle of artificial intelligence systems are consistent with obligations to protect human rights, as enshrined in applicable international law and in its domestic law”.
Fortifying nuclear assets
In a significant move to bolster cybersecurity in the nuclear sector, the UK Nuclear Decommissioning Authority (NDA) unveiled its Group Cyberspace Collaboration Centre, a state-of-the-art facility in Cumbria. The Centre marks a pivotal step in the NDA's strategy to enhance digital defences across its operations and will serve as a nexus for cross-sector collaboration, bringing together nuclear operators, cybersecurity experts and technology specialists.
Google launches London AI campus
Following a successful pilot, Google officially launched its AI Campus in Camden, London. The facility, inaugurated by Prime Minister Keir Starmer, is designed to provide local sixth-form students with a two-year pilot programme that offers access to cutting-edge resources in AI and machine learning. Collaborating with Camden Council and Camden Learning, the campus will equip students aged 16 to 18 with mentorship from industry experts, real-world project experience, and insights into how AI can address challenges across various sectors, including healthcare and the arts.
UK's AI Airlock: pioneering regulatory reform for AI-powered medical devices
The UK Medicines and Healthcare products Regulatory Agency (MHRA) launched the AI Airlock pilot scheme, selecting five innovative AI-powered medical technologies to participate in this groundbreaking regulatory sandbox. This initiative aims to revolutionise the regulation of artificial intelligence in medical devices, addressing the unique challenges posed by AI's evolving nature in healthcare applications. The pilot, running until April 2025, will explore regulatory frameworks that enable swift and safe development of AI medical devices, potentially transforming patient care in areas such as cancer diagnostics and chronic respiratory disease management. By collaborating with manufacturers, the NHS and other stakeholders, the MHRA seeks to refine its approach to AI regulation, ensuring innovative technologies reach patients quickly while maintaining rigorous safety standards. The findings from this pilot are expected to significantly influence future UK AI Medical Device guidance and regulatory processes, positioning the UK at the forefront of AI integration in healthcare.
North American initiatives
Google asks US Government to intervene in Microsoft and OpenAI deal
Google has reportedly appealed to the US Federal Trade Commission (FTC) to dissolve Microsoft's exclusive agreement with OpenAI, which allows only Microsoft to host the AI firm's technology on its Azure cloud platform. This partnership, valued at over $13 billion, has positioned Microsoft as a dominant player in the AI and cloud computing sectors, but Google argues that it stifles competition by forcing businesses to adopt Microsoft's services to avoid additional costs. The FTC is currently investigating Microsoft's business practices, and if it agrees with Google's request, this could open the door for other cloud providers, such as Google Cloud and Amazon Web Services, to host OpenAI's models, enhancing competition and potentially reducing costs for consumers.
Industry initiatives and events
Ori Industries first in the UK to deploy Nvidia H200 chips
Ori Industries is set to become the first company in the UK to deploy Nvidia's new H200 chips, a significant leap forward in AI computing capabilities. Starting in January 2025, the London-based tech firm will offer these powerful GPUs, which feature 141GB of memory and an impressive 4.8 terabytes per second bandwidth - nearly double that of the previous A100 model. Ori is now accepting applications for capacity reservations and plans to introduce Nvidia's GB200 exascale computers in April 2025. Mahdi Yahya, CEO and founder of Ori, confirmed that they will cater to both UK and international companies needing computer resources based in the UK. Founded in 2018 and backed by Episode Ventures and Force Over Mass Capital, Ori specialises in GPU cloud infrastructure for AI applications, positioning itself at the forefront of AI infrastructure in the UK and enhancing the country’s capabilities in supporting advanced AI research and applications.
Google unveils Willow quantum chip
Google unveiled its Willow quantum computing chip, marking a significant milestone in the quest for practical quantum computing with potential implications for AI. The chip demonstrated the ability to solve a complex computation in under five minutes that would take a supercomputer approximately 10 septillion years to complete. Willow's key innovation lies in its capacity to exponentially reduce errors while scaling up the number of qubits, addressing a long-standing challenge in quantum computing. While the immediate applications are primarily experimental, Google is exploring quantum algorithms to enhance foundational computational tasks for AI, including data processing and model training. The company suggests that quantum computers could eventually access training data currently beyond the reach of classical computers, potentially revolutionising AI development and capabilities in fields such as drug discovery and encryption.
Sam Altman's nuclear venture aims to power AI's future
Nuclear start-up Oklo, backed by OpenAI’s Sam Altman, has entered into a strategic partnership with data centre operator Switch, setting out plans for Oklo to construct small modular reactors (SMRs) capable of generating 12 gigawatts of electricity by 2044, potentially marking the largest corporate clean energy agreement to date. The partnership aims to address the surging energy demands of AI-driven data centres, serving tech giants like Google and Nvidia. While Oklo targets late 2027 for its first commercial reactor, pending regulatory approval, significant hurdles remain. The company must secure approval from the Nuclear Regulatory Commission after a previous setback in 2022, with plans to resubmit an application in 2025.
AWS re:Invent 2024
Amazon Web Services' flagship conference in Las Vegas saw the announcement of Amazon Nova, a new generation of state-of-the-art multimodal foundation models, one of several key innovations announced by CEO Andy Jassy. Amazon unveiled six new Nova foundation models, text models Micro, Lite, Pro and Premier (coming in Q1 2025), and creative models Canvas (image generation) and Reel (video generation). Nova models already power about 1,000 generative AI applications within Amazon. AWS also announced new Trn2 instances featuring AWS's newest Trainium2 AI chip, and Aurora DSQL, a new serverless distributed SQL database.
We expect AI adoption and implementation will continue to be a priority for businesses in 2025. The rapid evolution of AI capabilities brings unprecedented opportunities for innovation and efficiency, but also demands thoughtful consideration of legal and ethical implications. Successful AI integration will require a balanced approach: embracing technological advancement while maintaining robust governance frameworks and ensuring responsible deployment.
To discuss how we can support you on your AI journey, please get in touch with Tim Wright or Nathan Evans.
