Last week, Ofgem released guidance on the ethical use of artificial intelligence (AI) in the energy sector. Rather than introducing new regulation, it sets out a principles-led framework for deploying AI safely, fairly, and transparently across energy systems. The guidance touches on everything from governance and risk management to data practices, supply chain accountability, and organisational competency. 

For energy companies already investing in AI-powered systems, this marks a clear shift. Success is no longer just about technical performance. It’s about building trust, maintaining oversight, and embedding safeguards into digital operations from day one. 

At Hypercube, we’ve seen this shift unfolding on the ground. Many of the organisations we work with are moving beyond questions of technical feasibility. The focus now is on how to embed oversight, maintain explainability, and ensure AI systems remain reliable and accountable over time. Ofgem’s guidance offers a valuable structure for those conversations. Here’s our take on what it covers – and why it matters. 

Governance: Make AI accountable across the lifecycle 

Ofgem stresses the need for proportionate and effective governance. This includes board-level oversight, clear roles and responsibilities, and policies that cover the full AI lifecycle – design, development, deployment, monitoring, and decommissioning. 

Importantly, Ofgem recognises that not every organisation has the same level of maturity. Governance doesn’t need to be overly complex, but it must be proportionate to the risks involved. Smaller organisations can achieve this through streamlined processes and clear internal communication. For larger utilities or system operators, more formal oversight – such as a dedicated AI committee or cross-functional risk forum – may be necessary. 

At Hypercube, we’ve found that organisations often want to do the right thing, but lack a clear framework for how governance should look in practice. Ofgem’s emphasis on proportionate oversight aligns with how we approach system design – integrating accountability at every stage without overburdening delivery teams. 

Risk management: Proportionality and context over checklists 

AI introduces new forms of risk, not just because of the technology itself, but because of how it operates within broader systems. Ofgem highlights that AI risk assessment should focus on outcomes. This is a shift from traditional software assurance, which often relies on deterministic testing. AI models behave probabilistically. They learn from data, and their performance changes over time. That makes testing, validation, and context-awareness essential. 

The guidance recommends that organisations carefully assess whether AI is the right tool for the job. If it is, then a structured risk framework should follow. This includes defining what “good enough” looks like for each application, understanding where failures may occur, and identifying how recovery and mitigation will be handled. 

Ofgem’s call for continuous model review resonates strongly with our experience supporting AI deployment in live energy environments. We’re aware of how even well-performing models can drift without warning, especially when exposed to real-world conditions like asset downtime, seasonal behaviour shifts, or market volatility. Designing AI systems to respond to change isn’t optional, it’s foundational. 

Fairness and transparency: Beyond the algorithm 

AI has the potential to reinforce bias, particularly when training data does not accurately reflect the diversity of the population or user base. In the energy sector, this is a serious concern. AI tools are increasingly used to make decisions about customer support, service prioritisation, and billing structures. If these systems produce outcomes that are unfair, trust in both the system and the provider will erode quickly. 

Ofgem encourages organisations to treat fairness as a system-wide issue. It is not enough to audit an algorithm in isolation. The broader context matters, including how outputs are used, who is affected by decisions, and what avenues are available for recourse or redress. 

Transparency plays a key role in fairness. Stakeholders – including customers, regulators, and internal users – must understand how decisions are being made. This is particularly important in customer-facing applications, where explainability and the ability to challenge or override an AI recommendation can have real-world impacts. 

Sustainability: AI systems with energy impact in mind 

The guidance also addresses a growing concern: the energy and environmental cost of AI. This was a key theme uncovered at the UK All-Energy conference in Glasgow. Data centres, compute-intensive models, and large-scale training pipelines all consume significant amounts of electricity. As the sector works to decarbonise, it’s vital that digital tools support this goal rather than undermine it. We recently hosted Giles Hanglin, CEO of Apatura on the Hypercube Podcast – he breaks down the data centre opportunity in Scotland and Apatura’s perspective on how data centres and co-location are designed for energy efficiency:

Ofgem encourages organisations to consider the sustainability of their AI systems. This includes selecting infrastructure that aligns with low-carbon goals, using efficient model architectures, and avoiding over-engineered solutions. Where possible, energy consumption should be measured and reviewed alongside system performance. 

This part of the guidance sends a clear message: efficiency in AI is not just about speed or accuracy. It’s also about responsible resource use, especially when deployed at scale across national infrastructure. 

Security and supply chains: Guarding the expanding risk surface 

As AI systems grow more complex and interconnected, their attack surface expands. Ofgem urges energy organisations to build security in from the outset. This includes protecting training data, securing model outputs, and monitoring third-party systems involved in the supply chain. 

The guidance makes it clear that supply chain responsibility extends to AI. If a model is developed or hosted externally, the organisation using it must still ensure that it meets ethical and operational standards. This includes validating how data is handled, how updates are delivered, and whether the system can be audited if necessary. 

Competency: Why AI readiness depends on people 

Of all the sections in the guidance, Practice 7 on access to competent persons may be the most immediately actionable. Ofgem points out that the greatest risk with AI often lies not in the model, but in how it is used – or misused – by people who don’t fully understand it. 

This is particularly relevant in high-stakes environments like energy forecasting, asset management, and system control. AI models might surface complex patterns, but interpretation and operational judgment still lie with human users. 

Organisations are advised to invest in training, define role-specific knowledge requirements, and ensure that governance, technical, and operational teams are aligned in their understanding of how AI works. This includes creating clear documentation, building internal forums for sharing lessons learned, and planning for regular capability updates. 

This is an area we’ve leaned into heavily. Technical expertise alone isn’t enough; confidence in using AI comes from shared understanding between data teams, operational staff, and decision-makers. Some of our most impactful projects have started not with model training, but with aligning expectations and building internal fluency across roles. 

Use cases: AI in action requires practical design 

Ofgem closes the guidance with sector-specific examples, covering everything from predictive maintenance and demand forecasting to customer interaction and energy market trading. These use cases highlight the diversity of AI applications in energy – and the importance of tailoring governance and risk management to each. 

AI in forecasting, for example, requires careful model monitoring and fallback planning. AI in customer support needs clarity, explainability, and built-in protections against misinformation. And AI in trading raises entirely different challenges, including competition law and anti-collusion risk. 

The message is clear: ethical AI is not just about choosing the right model. It’s about designing the right system, in the right context, with the right people around it. 

Partnering for long-term success in AI initiatives 

Ofgem’s guidance sets a new benchmark for responsible AI use in the UK energy sector. It doesn’t prescribe every detail but provides a strong foundation for action. It also makes one thing clear: doing AI well is not just about deploying tools – it’s about building systems and teams that can adapt, learn, and respond to change. 

That’s why our role as AI partner for the energy sector is focused on long-term collaboration. We help energy organisations navigate the practical challenges of AI strategy, system design, capability-building, and lifecycle governance.