September saw a number of interesting announcements and developments in the field of artificial intelligence which we discuss in this month’s round-up.
EU launches consultation on Code of Practice for General-Purpose AI
The European Commission initiated a consultation process for developing a Code of Practice for General-Purpose AI models, as mandated by the EU AI Act. The code will address critical areas such as transparency, copyright rules, risk assessment and mitigation for GPAI systems. The Commission is seeking input from GPAI providers, businesses, civil society and academic experts to inform the draft code. Additionally, the AI Office is developing guidelines for summarising training data used in GPAI models. The Commission aims to finalise the Code of Practice by April 2025, with GPAI provisions of the AI Act set to apply from 1 August 2025.
AI governance initiatives
The UN Secretary-General's High-level Advisory Body on Artificial Intelligence released its final report, "Governing AI for Humanity," which outlines recommendations for global AI governance, including:
- establishing an International Scientific Panel on AI;
- launching a new policy dialogue on AI governance at the UN;
- setting up an AI standards exchange;
- creating a global AI capacity development network;
- establishing a global AI fund;
- fostering a global AI data framework; and
- forming an AI office within the UN Secretariat.
UK Lords propose bill to regulate AI in the public sector
A new private member's bill introduced by Lord Clement-Jones in the House of Lords aims to enhance regulation of AI systems used by UK public authorities. The proposed legislation seeks to ensure greater transparency and accountability in AI deployment within the public sector. Dubbed the ‘computer says no’ bill, it would require public bodies to disclose when AI is used in decision-making processes affecting individuals, addressing concerns about potential biases and lack of human oversight in AI-driven public services, and emphasising the need for explainable AI and clear accountability mechanisms in government AI applications. However, as a Private Member's Bill, its chances of becoming law are limited.
Australian government announces new AI safety regulations
The Australian government announced two key measures to enhance AI safety and regulation:
- a Proposals Paper which outlines plans for mandatory guardrails for high-risk AI applications, including a proposed definition of high-risk AI and ten mandatory standards, with options for regulatory approaches; and
- a Voluntary AI Safety Standard which provides immediate guidance to businesses using high-risk AI, ensuring clarity before mandatory regulations are implemented.
Public consultation on the Proposals Paper closes on 4 October.
Global consensus emerges on military AI governance
At the recent Responsible AI in the Military Domain (REAIM) summit in Seoul, nearly 100 countries endorsed a 'blueprint for action' to govern the responsible use of AI in military applications. This significant agreement aims to establish a framework for the ethical development and deployment of AI in defence contexts, and states that humans - not AI - should make key decisions regarding the use of nuclear weapons. Notably, China refused to sign the agreement.
European Parliament calls for expanded AI Liability Directive
A study commissioned by the European Parliament's Committee on Legal Affairs recommended a significant expansion to the scope of the proposed AI Liability Directive, to include general-purpose AI systems and extending liability to cover software more broadly. If adopted, these changes could significantly impact the landscape of AI regulation and liability in the EU.
Regulators clear Microsoft-Inflection AI merger
The UK's Competition and Markets Authority has concluded its investigation into Microsoft's hiring of key personnel from Inflection AI, determining that the transaction qualifies as a merger but poses no significant competition concerns. The CMA found that while Microsoft effectively acquired nearly all of Inflection's team, including two co-founders (often referred to as an aqui-hire), this does not substantially lessen competition in the market for consumer chatbots. The European Commission also announced that it is dropping its own investigation into the transaction.
Industry announcements and events
InfoQ's AI, ML and Data Engineering Trends Report for September 2024 highlighted several key developments, reflecting significant progress in AI investment, governance, academic research and industry applications, including:
- increased importance of Retrieval Augmented Generation (RAG) for large-scale LLM applications;
- growing attention to AI-powered hardware, including AI-enabled GPU infrastructure and AI-powered PCs;
- exploration and adoption of small language models (SLMs) for edge computing use cases; and
- advancements in AI agents, particularly coding assistants in corporate application development settings.
BlackRock, Global Infrastructure Partners, Microsoft and MGX announced the Global AI Infrastructure Investment Partnership, a fund which aims to invest up to $100 billion in new and expanded data centres and supporting energy infrastructure, primarily in the United States, to fuel AI innovation and economic growth.
LinkedIn updated its privacy policy for European users, clarifying that it does not use data from members in the UK, EU, EEA or Switzerland to train generative AI models. For non-EU users, a default setting automatically authorises this data usage. LinkedIn’s action follows similar moves by other tech companies. X suspended processing personal data from EU users for AI training in August, and Meta is reported to have halted its AI rollout in Europe after requests from data protection authorities.
Alibaba Cloud announced the release of over 100 open-source large language models as part of its Qwen 2.5 series, aiming to boost global AI development. These models, ranging from 0.5 to 72 billion parameters and supporting 29+ languages, cater to various industries including automotive, gaming and scientific research. The release is part of Alibaba's strategy to invest heavily in AI technology and infrastructure, attracting global customers to its cloud services.
Microsoft is reported to have struck a ground-breaking 20-year deal to purchase all power from the reopened Three Mile Island nuclear plant, aiming to secure zero-emissions electricity for its AI operations. The plant, which closed in 2019, is set to reopen by 2028 following a $1.6 billion investment, partly supported by tax breaks from the 2022 Inflation Reduction Act. This is the first time a decommissioned US nuclear plant will resume operations and the first instance of a single commercial customer claiming a nuclear plant's entire output.
IBA and CAIDP release report on AI in the legal profession
And last, but by no means least, the International Bar Association and the Center for AI and Digital Policy jointly published a landmark report titled "The Future is Now: Artificial Intelligence and the Legal Profession." The report highlights the transformative potential of AI in legal practice while emphasising the need for responsible adoption, and calls for the legal community to proactively embrace AI while ensuring fairness, accountability and transparency. It serves as a crucial resource for legal professionals and policymakers navigating the integration of AI technologies in the legal sector.
As the AI revolution unfolds, businesses face the challenge of navigating an increasingly complex regulatory landscape. To explore how we can support you on this journey, please get in touch with Tim Wright or Nathan Evans.
