ComplyAdvantage

Clarity on how to govern AI effectively in a way that doesn't slow down progress. Mandatory encoding of human rights into AI development. Clarity on a governance approach that enables safe adoption of AI in more use cases, through increased trust and explainability.

• Safe, secure and trustworthy AI
• Transparency, accountability, and human oversight
• Protection and promotion of human rights
• Social, economic, ethical, cultural, linguistic and technical implications of AI

If AI isn't seen as safe and trustworthy, it will not be adopted as widely as it could be. The tech is coming on, but adoption lags. People are concerned about its impact on humanity and whether it can be trusted to do the job consistently. Most people accept it works, but lack confidence. AI Governance os critical to bridge this trust gap so we can realise the benefits fully.

Human adaptation. This is partly for people to change how they work or interact with AI as a customer, but also as a society. Are we ready to the impacts of AI on society and the tensions it will create? AI is used by criminals to commit crime, bad actors to influence public opinion through social media and will also change the jobs market. This is a bad cocktail and it is up to those implementing AI to consider their impact on the whole. If everyone acts in their best interests, the whole will suffer. We can't simply ignore these impacts when launching new AI capabilities. All of us in the AI governance community need to work on the shared problem of how humans adapt and how to minimise negative consequences.

The main issue in AML/Sanctions sector is a tension between those who see AI saving money through automation and those who are concerned about whether it can be trusted, or what happens to the data fed to it. Some of this is a knowledge gap (both in terms of what it is actually capable of and how concerned we should really be by it), but it falls to AI Governance to bridge this gap often. We are also the ethical conscience of developers who might not always see beyond the technical problem solving. The main gap is knowledge & trust of the public and consumers. Education is important, too much of the conversation is owned by frontier model developers who have conflicts of interest in terms of helping people understand the impacts/capabilities/risks associated with AI. We need a balanced and unbiased voice to help people understand AI better

Setting standards of governance, helping to identify risks and controls, outlining expectations around oversight, being an unbiased voice on AI and risks.

ISO42001 is a good foundation. We need more global standardisation and clarity on how to oversee AI. Partnerships are helpful, but we can't allow for regulatory capture, given vested interests in the industry. We need to bring users into the dialogue, not just AI producers. Users can be corporate/gov't consumers of AI models as well as public/charity representatives

There needs to be the ability for a dynamic community, rather than rigid updates that do not allow for timely exchanges. Things are moving quickly and the Dialogue needs to be able to keep up, or it will become irrelevant. The community can be subdivided according to industry and also AI role (producer of model or integrator of models). These can be mixed up at times, but in general there needs to be subdivisions to maintain dialogue. Also actions need to be taken, not just dialogue; how are we going to move it forward, not just spectate on the issues?

There are definitely gender, geographic and ethnicity gaps, but there are more subtle gaps too. Often it comes down to familiarity with the topic. I've attended roundtables on this where there are huge differences in experience despite many similarities on paper (same industry, same education, geography, ethnicity and gender). These segmentations alone will not account for how close people are to AI and how able they are to understand and govern its impacts. Likewise, I don't think these typical labels are good predictors alone in understanding how familiar people will be with AI. Clearly less economically developed countries will have less exposure to AI, but within economies and industry segments, there are more subtle delimiters. These can include whether one's company is engaged with AI (and therefore there are opportunities to learn more) or whether one can adopt AI into one's personal life. It can be very hard to predict affinity to AI and understanding of AI impacts, but looking at a person on paper.

a mixture of groupings that allow for detailed discussion and also challenge. Therefore, sometimes create smaller groups where people have similar experiences and other times mix so that people's ideas are challenged. Involve model producers, but also consumers and public bodies to drive creative discussion and challenge of assumptions. Hackathons are also hugely engaging. Tasks can be submitted and completed with scoring and prizes. Engagement comes through action and results. Anything that's a talking shop will be ignored or bypassed by a fast moving industry. The Dialogue will need to galvanise action and be a source that people look to for practical guidance, so it must have the best thinkers involved and put out practical steps, not dense theory. The dialogue needs to unify national and super-national bodies so that we end the proliferation of marginally different regulatory regimes (EU AI Act, OECD Guidance, governmental guidance etc). This becomes a complex and overlapping web of requirements for a tech that spans the globe. We need the dialogue to help people, not get in their way. If it simplifies and unifies and provides practical help, people will engage with it

ISO42001 and the UK white paper: a pro innovation approach to ai regulation are practical examples. OCC Guidance on MRM (pre April 2026 update) was also very helpful. Overall, the best way for governance to be effective is to complement the task it is governing, not be a blocker. In this industry in particular that is most true. In my role, AI governance has become a seamless part of product development, with very low friction or overhead, with strong emphasis on product quality as well as compliance. This has meant that people conform and we have seen a huge improvement in product as well as explainability to our clients. AI Governance should not be a blocker or reason not to try something, it should be there to ensure it's done safely and effectively. Ungoverned products are likely to not be good products, not just ungoverned. Establishing that link between oversight and outcome is the key to an embedded and effective governance regime, particularly with a dynamic field like AI. You don't need flashy tech or even complex frameworks, you just need to align oversight with business need and ensure stakeholders are bought in; the rest follows.