AI Safety

What do we expect from the AI Action Summit

The AI Safety Institute’s Ambitions and Progress Reports

How the AI Safety Institute is Approaching Evaluations

Fourth Progress Report Towards Ambitions of AI Safety Institute

Third Progress Report Towards Ambitions of AI Safety Institute 

Second Progress Report Towards Ambitions of AI Safety Institute

First Progress Report Towards Ambitions of AI Safety Institute

The UK and US Announce a Partnership to Safety Test AI 

UK Government to partner on AI Safety with The Republic of Korea

DSIT Releases the interim International Scientific Report on the Safety of Advanced AI

UK Launches £200,000 Grants for Systemic AI Safety: Fostering Trust and Safety

techUK is thrilled to commence the AI Action Summit events in Paris, with our first day proving to be full of productive international dialogue and collaboration around AI governance and innovation. 

The day commenced with significant techUK engagement across several key events in Paris. At the British Embassy, techUK supported an event titled "AI for Growth: Fostering UK-France Collaboration." This gathering brought together leaders from both nations' AI ecosystems, facilitating deep discussions through focused roundtables. The sessions explored AI adoption opportunities across Financial and Professional Services, Life Sciences, and Telecommunications sectors, while also addressing fundamental aspects of the current AI landscape including computing infrastructure, skills development, and sustainability considerations. The event featured notable speakers including UK AI Minister Feryal Clark MP, UK Ambassador Dame Menna Rawlings, Industrial Strategy Advisory Council Chair Clare Barclay, and Mistral AI's Global Head of Public Affairs Audrey Herblin-Stoop.

As a founding member of the Tech7, TechUK's presence in Paris alongside sister organisations from G7 countries reinforced the commitment to international cooperation in AI governance. The day began with an informal dinner, setting the stage for policy discussions to follow and the upcoming AI Action Summit. While no major announcements were made during these events, there were indications of forthcoming French initiatives, particularly regarding data center developments. techUK enjoyed attending an insightful talk by Thomas Courbe, Director General for French DG for Enterprise and France's representative on the EU AI Board - the body established by the EU AI Act to facilitate coordination and discussions between member states.

The day also saw TechUK's participation in the inaugural conference of safe and ethical AI, hosted by IASEAI. This newly-formed independent charitable organisation brought together approximately 700 in-person attendees from 65 countries, representing academia, industry, government, and civil society. The conference successfully worked to break down existing silos in AI risk mitigation, featuring influential speakers such as Yoshua Bengio, Geoffrey Hinton, Margaret Mitchel, Ancan Dragon, and Maria Ressa. The event demonstrated a unified commitment to ensuring AI systems operate safely and ethically, while shaping policy and promoting research in this crucial field.

The overall atmosphere across all events was notably positive, establishing a constructive foundation for upcoming discussions and collaborations in the AI space. This day of engagement underscores the growing momentum behind international cooperation in AI governance and the shared commitment to responsible AI development and implementation.

Catch up with all the content here.

A key outcome of today was the OECD's launch of the first global framework for AI reporting. Leading technology companies including Microsoft, NTT, Safer AI, NEC Corporation, OpenText, and Google are working closely with the OECD on this framework, which monitors the application of the Hiroshima Process International Code of Conduct for Organisations Developing Advanced AI Systems.

The second day of the AI Action Summit featured significant policy discussions with key international stakeholders.At the Tech7 the UK AI Minister Feryal Clark MP delivered an insightful keynote, followed by a comprehensive presentation from Thomas Courbe, with active participation from G7 embassy representatives. During these discussions, our CEO Julian David spoke on a panel and emphasised the evolution of the Hiroshima process from the G7 to the OECD and its crucial role in establishing principles for global AI cooperation. This builds on Tech7's foundation, which was initiated by techUK in 2021 to unite digital sector voices across G7 countries and the EU in advocating for international cooperation on technology and trade.

Later in the day, techUK participated in a panel at Numeum's "Electrify the Now of AI: Together We Create" event. This gathering featured distinguished speakers including Courtney Lang, Vice President of Policy, Trust and Technology at ITI, Tech7 representatives, and French Minister for AI and Digital Technologies Clara Chapaz. The discussions painted an optimistic picture of AI's future and the potential for international collaboration.

The second day of the IASEAI conference brought forward crucial perspectives on AI safety and sustainability. Kate Crawford presented a compelling analysis of AI's environmental impact, outlining a comprehensive sustainability framework that bridges AI ethics and safety. Her presentation highlighted four key environmental challenges: mineral consumption, energy usage, water consumption, and e-waste generation. The conference featured MH Tessler's innovative work on democratic deliberation and information access, introducing new approaches to evaluating machine-generated content through human participation. Looking toward 2025, participants outlined practical steps including accident prevention strategies, model registration requirements, and context-specific robustness measures. n his closing remarks, Stuart Russell offered a striking metaphor: AI is like an airplane with all of humanity as its passengers - we must keep it flying and once it takes off we can never allow it to crash. The metaphor underscores why responsible AI development is critical.

The eve of the France AI Action Summit saw Paris buzzing with multiple high-profile events across the city, showcasing the growing momentum behind international AI cooperation.

The day kicked off with a visit to the AI and Society House organised by Humane Intelligence, led by Rumman Chowdhury. The US-based tech non-profit is focused on building a community of algorithmic assessment and test and evaluation programmes focused on the societal impact of AI. The packed event saw representatives from NGOs, industry, civil society and policy from around the world with panels discussing issues including industry-civil society-government collaboration and demos showcasing global projects focusing on bias and red teaming projects currently underway.

The evening featured a series of enriching community events across Paris, including Google Art and Culture's inspiring showcase of AI's creative potential, an impactful Women in AI event that brought together leading voices emphasising the importance of diverse communities in shaping AI's future, and Cohere's reception which provided valuable industry networking opportunities.

A highlight of the day was the comprehensive program "Governing in the Age of AI" hosted by the Tony Blair Institute for Global Change. The event opened with remarks from former Italian Prime Minister Matteo Renzi and Anne Bouverot, Special Envoy of the French President to the AI Action Summit. Key panels explored global leadership and AI opportunities, with Matt Clifford highlighting AI's role as an engine of growth and key aims of the AI Opportunities Action Plan. TechUK members Faculty and AMD contributed valuable insights, emphasising that there's no difference between economic, energy, industrial and AI strategy - they're all interlinked. Distinguished scientist Yoshua Bengio highlighted the importance of evidence-based policymaking while acknowledging current uncertainties in the field.

At the Canadian Embassy, the Ambassador and Canadian Minister of Digital emphasised the importance of transatlantic cooperation in AI governance, highlighting Tech7's supporting role. A significant announcement garnered attention with Brookfield's commitment to invest $20 billion in French AI infrastructure, demonstrating the growing international investment in AI capabilities. 

The discussions set the stage for tomorrow's AI Action Summit, with emphasis on balancing innovation with responsible development. Key themes emerged around sustainable AI, energy considerations, and the framework for global governance, highlighting the growing alignment between industry, civil society, and government stakeholders.

The AI Action Summit officially opened at the iconic Grand Palais, bringing together over 800 contributors from more than 100 countries. Special Envoy Anne Bouverot emphasised three crucial points in her opening address: the importance of global openness, the need to move from science fiction to real-world applications, and the urgency to act for shared progress in AI development. Citing Marie Curie, she reminded attendees that understanding, not fear, should guide our approach to AI.

Dr. Fei-Fei Li, often referred to as the "godmother of AI," delivered a compelling keynote, emphasising that "artificial intelligence, in reality, is not artificial." She praised France's position as a major player in AI research and the Summit's action-oriented approach. The morning also featured an insightful panel with META FAIR Director Joelle Pineau and former Metamate Fidji Simo, exploring dataset challenges in advancing AI for scientific progress.

A significant parallel IAPP AI Literacy Event hosted by Hogan Lovells brought together key figures from industry and regulation for discussions on AI literacy. The discussions centered on the critical importance of transparency in AI systems, focusing on model operations, data usage, and helping both public and industry stakeholders better understand AI's capabilities and limitations.

The day concluded with TechUK CEO Julian David addressing the US Chamber of Commerce dinner. His remarks struck an optimistic tone about AI's potential, highlighting the UK's balanced approach between safety principles and innovation. He emphasised the importance of international dialogue on safety while ensuring AI benefits reach across society, economy, and environmental sustainability.

The day's events underscore the growing global consensus on the need for balanced AI development - maintaining safety while advancing innovation and practical applications. The diversity of perspectives and depth of discussions continued.

The momentum of the AI Action Summit continued with a dynamic series of parallel events across Paris and London, demonstrating the truly international nature of the ongoing dialogue on AI development and governance.  

The day began with TechUK's Head of International Policy, Sabina Ciofu, attending and AWS breakfast before speaking at the AI Fringe panel in Paris on "Delivering Trustworthy AI in Challenging Times." The discussion emphasised data flows, regulatory cooperation, and the UK's progress in developing practical tools for operationalising ethics through assurance mechanisms and standards. A key point was the need for an international common language on transparency that serves customers, regulators, and investors alike. 

Station F hosted crucial discussions on the challenges of resource competition in AI development, exploring the distribution of computing power, talent, and data resources. The sessions highlighted the stark contrast between US tech giants investing $40-80bn each in AI development versus European companies like Mistral raising $1bn. The dialogue also addressed the role of competition authorities and the complexities of regulatory compliance, particularly for smaller companies facing disproportionate burdens compared to established players. 

At the International Chamber of Commerce (ICC), Datasphere unveiled its report "Sandboxes for AI: tools for a new frontier." The event brought together an impressive panel including representatives from Microsoft, OpenMined, Japan's Ministry of International Affairs, the Alan Turing Institute, OECD, and Singapore's IMDA. Discussions emphasised the importance of adaptive frameworks and the Hiroshima process in shaping international AI governance. 

Meanwhile in London, at the British Library techUK hosted two significant AI Fringe panel sessions moderated by Sue Daley OBE. The first explored "The Industrialisation of AI," examining how we can support AI adoption across the economy. The second session focused on "AI in Public Services," investigating practical implementation strategies in the public sector and the importance of trust and data to get this right.  

From left to right: Peter Corpe, Josh Robson, Rocio Concha Galguera, Kanishka Narayan MP, Sue Daley OBE 

From left to right: Sue Daley OBE, Kirsty Innes, Sir Nigel Shadbolt, Duncan Brown, Imogen Parker, Paul Maltby

In her closing remarks at London's AI Fringe gathering, Minister Feryal Clark highlighted how the AI Opportunities Plan will unlock AI's benefits, with aims to make compute 20x more accessible by 2030 and foster local AI growth zones. She noted that the government have launched the express of interest for AI growth zones. Feryal emphasised inclusive growth through initiatives to diversify the talent pool, ensure SME representation, and announced an exciting new AI Energy Council to collaborate with energy companies on clean energy solutions. The Minister also noted the success pre-summit engagements of Tech7 meetings and TechUK-supported events at the British Embassy in strengthening international partnerships. 

TechUK's leadership continued to play a key role in Paris, with CEO Julian David participating in a panel at Station F on "National AI Roadmaps: A Round Table on International Cooperation," while Sabina Ciofu contributed valuable insights at an Atlantic Council roundtable lunch discussion. 

The day concluded with a series of well-attended industry receptions hosted by Anthropic, Mozilla, and AWS, providing valuable networking opportunities and continued dialogue. These events underscored the day's key themes: the delicate balance between innovation and regulation, the critical importance of international cooperation, and the need for frameworks that support both established players and newcomers in the AI ecosystem. 

As we reflect on another remarkable (and final) day of the AI Action Summit, it's clear that the global AI community is moving from conversation to concrete action. The parallel discussions in Paris and London demonstrate not just the scale of collaboration needed, but also the shared commitment to ensuring AI benefits reach across society, economy, and our planet.  

Tomorrow brings another day of opportunities to shape this transformative technology together – but for now this insight series comes to a close, thanks for tuning in.  

For more information about techUK’s Digital Ethics and AI Safety Programming please contact [email protected]. 

Outcomes

So what were the key outcomes of the inaugural Global AI Safety Summit:  

• The Bletchley Declaration: signed by 28 countries, including the USA, China and European Union the Bletchley Declaration recognises that if the opportunities of AI are to be seized there must be an international effort to research, understand and mitigate the risks posed by frontier AI technologies.  

• More AI Safety Summits: the Bletchley Declaration confirmed additional meetings in 2024 with South Korea to host a mini virtual summit in six months while France will host the next full in person AI Safety Summit 12 months from now.  

• AI Safety Institute: The UK announced that it will put its Frontier Models Taskforce on a permanent footing in the form of a New AI Safety Institute creating a UK based but internationally facing resource with the purpose of evaluating frontier systems, advancing research on AI Safety and sharing information between a global network of Government’s, private companies and civil society organisations. The Institute was announced alongside endorsements from the US, Singaporean, German, Canadian and Japanese Governments as well as from major frontier AI labs.  

• Senior government representatives from leading AI nations, and major AI organisations, agreed to a plan for safety testing of frontier AI models: The plan involves testing models both pre- and post-deployment, and a role for governments in testing, particularly for critical national security, safety and society harms.  

• The UK unites with global partners to accelerate development in world’s poorest countries using AI: UK and partners to fund safe and responsible AI projects for development around the world, beginning in Africa, with £80 million collaboration. 

• Investment in the ‘AI Research Resource’ for the AI Safety Institute: The investment into the AI Research Resource has been tripled to £300 million, up from £100 million (announced in March 2023), in a bid to further boost UK AI capabilities. The investment will connect Isambard-AI (based at Bristol University) to a newly announced Cambridge supercomputer called ‘Dawn’. Connecting these two supercomputers will give researchers access to resources with more than 30-times the capacity of the UK’s current largest public AI computing tools.  

Overall, both Rishi Sunak and Michelle Donelan, the Science and Technology Secretary, who led the first day of the summit, will be happy with what the UK has achieved. Securing a consensus on a process for approaching frontier AI technologies and establishing a new track in global AI discussions.  

However, as always, the proof will be in how these initial agreements and forums develop in the years ahead and if they can lead to tangible progress between countries who often have differing views.