Key Takeaways from the 2025 Oxford Generative AI Summit

On October 16-17, 2025, over 200 leaders from industry, government, academia, the media, and the law came together at Jesus College, Oxford for the 2025 Oxford Generative AI Summit (OxGen AI 25).  During the two-day summit, there were over 25 panels, keynote presentations, and fireside chats, covering a wide range of topics and perspectives.  Womble Bond Dickinson sent six representatives to the conference, and below are key insights that we gleaned.

Overall, three main themes emerged from the discussions:  change management, AI sovereignty (or national economic strategy), and trust and truth.

Change Management:  Change management best practices are more essential than ever in the age of AI:

• Effective enterprise-wide AI adoption requires that even those with the lowest levels of tech skills or enthusiasm be encouraged to embrace the new technologies – in ways that work for them.
• Don’t allow AI vendors to redefine your workflows. Be cautious before adopting a top down approach.  Optimizing AI solutions requires collaboration, dialogue and interaction between your internal stakeholders.
• Give permission to stakeholders to experiment, experience failure, and then revise and refine.  AI is too disruptive and too fast-moving to be perfectly honed out-of-the-box.
• There was a broad consensus that AI will result in tectonic workforce displacement and disproportionately impact early-career workers.  The rise of Agentic AI will further accelerate these trends.

The Problem of AI Sovereignty:  The AI economy and infrastructure are largely dominated by the US and China, each of which is home to “big tech” multinationals leading the AI revolution. This centralized ecosystem creates a challenge for the EU, the UK, and other nations and regions: What resources do they invest to compete with big tech? What opportunities can be created to partner with industry giants so that they can successfully participate in the AI transformation? How do they continue to foster innovation when foreign companies and institutions are siphoning some of their most talented citizens? And if massive investments or infrastructure changes are required, at what cost to other national priorities, whether cultural, environmental, or health related? 

Truth and Trust? The issue of AI hallucinations by LLMs is still with us.  OxGen speakers likened AI tools to “drunken graduate students” or postulated that AI can at best provide only “directional insight.” In other words, accuracy, or grounding AI outputs in reality, is not guaranteed – and leads to complex challenges:

• The industry is still asking what could cause LLM outputs to veer away from reality and truth.
• The ever-improving quality of AI-generated multimedia outputs is making it increasingly difficult to discern what is “real” and what is not.  AI generated media is increasingly indistinguishable from authentic ones.
• This also gives rise to many new questions about intellectual property, authorship and ownership – in both inputs to AI as training data, as well as outputs from AI.  Many of these new questions bend our existing IP concepts, sometimes to the point of breaking, where the existing rules cannot accommodate the new situations.
• The concern about hallucinations has led many providers to add citations and other feature sets in an effort to anchor outputs more faithfully to what is documented as true.
• Opportunities for bad actors to undermine public faith in what users see, experience, and believe to be true are growing exponentially.

Given our strong commitment to transatlantic growth of the digital sector, including AI, Womble Bond Dickinson was proud to sponsor OxGen AI 25 and to be well-represented at the conference.  In the US, Womble Bond Dickinson was recently named by BTI Consulting as one of the top 32 Generative AI litigation law firms.  We can help guide you through the thicket of legal and organizational challenges in the new age of AI.