Energy companies could receive new guidance on how to test and evaluate artificial intelligence systems under proposals from Ofgem intended to improve confidence in the technology’s use across the sector.
The Office of Gas and Electricity Markets (Ofgem) has opened a consultation on voluntary AI assurance guidance for businesses operating in Great Britain’s energy market.
The proposed approach would encourage organisations to develop evidence that their AI systems are safe, secure, fair, and effective for the purposes in which they are deployed.
It would not impose additional statutory duties, create a mandatory certification programme, or constitute regulatory approval of an individual AI system.
Instead, the guidance would complement existing energy regulation and Ofgem’s ethical AI guidance, providing a framework for assessing whether the controls applied to AI are appropriate to the risks involved.
The consultation opened on 18 November and closes on 15 January 2027. It follows an earlier call for input examining AI assurance across the energy sector.
Ofgem is inviting responses from energy suppliers, network operators, market participants, technology developers, consumer groups, assurance specialists, and organisations involved in delivering essential services.
Evidence behind automated systems —
Artificial intelligence is being developed for applications including energy forecasting, network planning, asset management, and customer service.
Those activities involve different operational risks. A tool supporting an internal administrative task may require a different level of evaluation from a system informing infrastructure decisions or customer outcomes.
Ofgem’s proposed approach is therefore outcomes-based, with assurance activity intended to reflect the intended application and the potential consequences of failure.
For an energy business, assurance can involve examining how an AI system was developed, what data it uses, how its performance is tested, and whether its limitations are understood.
Evaluation may also need to establish whether security and governance controls remain effective when a system interacts with established operational processes.
These questions can arise when organisations develop AI internally or acquire technology from specialist suppliers.
A company procuring an AI service needs sufficient information to assess its suitability for the proposed task. Documentation about accuracy, testing conditions, access controls, and known limitations can help support that assessment.
However, supplier assurances do not remove the purchasing organisation’s responsibility for meeting applicable regulatory requirements.
Existing obligations under energy regulation, data protection law, and relevant cyber security requirements continue to apply regardless of whether a business adopts voluntary assurance guidance.
Ofgem’s proposals do not replace those obligations. They seek to provide a more consistent approach to demonstrating that AI systems have been evaluated appropriately.
The regulator is also considering how assurance could support innovation without creating unnecessary burdens for organisations experimenting with lower-risk applications.
Applying identical procedures to every AI system could create disproportionate costs, particularly for smaller businesses and projects involving limited operational exposure.
An approach linked to risk and intended use would allow companies to determine the evidence needed for different applications while working towards common principles.
The consultation also examines how Ofgem’s guidance should relate to existing assurance standards and other regulatory initiatives, reducing the risk of businesses having to repeat similar assessments.
Two online stakeholder workshops are scheduled for 7 December 2026 and 6 January 2027 to gather feedback on the proposals.
Ofgem will use the consultation responses to determine the scope and content of any guidance it subsequently develops.
For now, the regulator is seeking a common approach to evaluating AI within the energy sector, while leaving existing legal duties and formal regulatory approval arrangements unchanged.




