Ofcom on AI in telecoms customer service: the General Conditions still apply
In short: A phone or broadband provider that puts AI in front of its consumer and small business customers must already meet Ofcom’s complaints rules in General Condition C4 and, for consumers, its vulnerable-customer rules in General Condition C5, and UK AI telecoms regulation needs no new rules at present (Ofcom discussion paper, 25 September 2026). Customers should be able to tell that they are dealing with generative AI, and AI tools should not create barriers to the support channels a provider must offer (Ofcom expectations, 25 September 2026).
A broadband or mobile provider that answers its consumer and small business customers through an AI chatbot is bound by the same rules as its call centre. It must still let those customers complain by telephone, post and email or web form, must still follow its own policies for consumers in vulnerable circumstances, and remains responsible for the chatbot even where another company built it. Those points are in a discussion paper on AI in telecoms markets that Ofcom published on 25 September 2026, and Ofcom is not intervening with new rules or guidance at present (paragraphs 1.16 and 6.19).
This post covers AI that deals with customers in sales, service, complaints and support. AI inside the network carries separate security duties, discussed in our post on Ofcom’s AI strategy for 2026/27.
Evidence on AI in telecoms customer service
There is limited evidence that AI tools have been a systemic cause of harm to telecoms customers, and Ofcom considers that its existing rules and framework enable it to protect consumers effectively (discussion paper, paragraph 1.16). In Ofcom’s Technology Tracker, 53% of adult UK internet users had interacted with a generative AI tool or an AI customer service chatbot in the previous 12 months. Of online adults, 8% used an AI tool in relation to their telecoms services and 2% used one to make a complaint. Many people in Ofcom’s qualitative research still wanted to speak to a person about complex or sensitive matters, while some preferred not to.
Ofcom expects providers, and anyone developing AI tools for them, to make it clear from how a generative AI tool describes itself that it is AI, and not to let a provider’s AI tools act as “gatekeepers” that create barriers to the support channels the provider must already offer. Both clarify rules and expectations that already apply (discussion paper, paragraph 1.17).
AI telecoms regulation: the General Conditions that apply
Ofcom sets the General Conditions of Entitlement (the conditions that apply to providers of electronic communications networks and services) under section 45 of the Communications Act 2003. The General Conditions apply to the AI tools providers adopt, within each condition’s own scope (discussion paper, paragraph 6.19). A provider that deploys an AI application designed by a third party remains responsible for that application’s compliance.
A provider of public electronic communications services to consumers, microenterprise and small enterprise customers and not-for-profit customers must handle their complaints under procedures that conform with Section 1 of the Ofcom Approved Complaints Code (General Condition C4.2(a)). Under paragraph 4 of Section 1, the provider must allow those customers to complain by at least three means: a free or geographic-rate telephone number, a UK postal address, and an email address or dedicated web form. Under paragraph 2, its complaints procedures must be accessible to customers with disabilities and customers in vulnerable circumstances. A chatbot can be a further route. A provider that does not keep all three routes open to those customers does not conform with the Code and so contravenes C4.2(a). Complaints from larger business customers fall outside General Condition C4.
A provider of public electronic communications services must also establish, publish and comply with clear and effective policies for the fair treatment of consumers whose circumstances may make them vulnerable (General Condition C5.2), including the channels through which those consumers can contact it (C5.3(b)). Ofcom expects a chatbot that deals with those consumers to follow the provider’s policies and to be only one of the channels open to them, alongside human contact by telephone, email, post and video and text relay services. Under C5.6 the provider must take the measures needed to meet the needs of end-users with disabilities, and Ofcom expects AI tools to meet those needs as a website or telephone line would.
| Customer-facing AI use | Rule that applies | What the provider must do |
|---|---|---|
| AI chatbot taking complaints from consumers and small business customers | General Condition C4.2(a) and the Ofcom Approved Complaints Code, Section 1 | Keep the telephone, postal and email or web form routes open and accessible |
| AI dealing with consumers in vulnerable circumstances | General Condition C5.2 to C5.5 | Make the AI tool follow the provider’s published vulnerable-customer policies (Ofcom expects human channels to stay available) |
| AI used by end-users with disabilities | General Condition C5.6 | Meet the needs of end-users with disabilities through the AI channel |
| AI tool built by a third party | The General Conditions that apply to the channel | Ensure the tool complies, as the provider that deploys it |
| AI taking a significant decision with no meaningful human involvement | UK GDPR Articles 22A to 22C | Give information, allow representations, offer human intervention and allow the decision to be contested |
Data protection rules for AI decisions about telecoms customers
UK GDPR applies alongside the General Conditions whenever a provider’s AI tool processes customers’ personal data. A decision is based solely on automated processing where there is no meaningful human involvement in it, and is significant where it produces a legal effect, or a similarly significant effect, for the customer (Article 22A of the UK GDPR). Where a provider, as data controller, takes such a decision on personal data, it must have safeguards that give the customer information about the decision and enable the customer to make representations, obtain human intervention and contest it (Article 22C). A provider may not take such a decision solely by automated means on special category data, such as health data, unless a narrow condition in Article 22B is met, such as the customer’s explicit consent. Section 80 of the Data (Use and Access) Act 2025 substituted these Articles for the former Article 22, and they have applied in full since 5 February 2026.
Telecoms providers are generally cautious with customer-facing generative AI, limiting its autonomous decision-making and keeping routes to human review (Ofcom discussion paper, paragraph 6.6). A provider that lets an AI tool decide, for example, whether to offer a customer a contract, with no meaningful human involvement, must consider whether the decision has a legal or similarly significant effect. Where it does, the Article 22C safeguards apply in addition to General Conditions C4 and C5. A provider using AI agents should also consider the ICO’s work on automated decision-making, discussed in our post on agentic AI data protection.
Complying with AI telecoms regulation: what providers must do
A provider needs no Ofcom approval to deploy customer-facing AI, and Ofcom has made no AI-specific rule. It still has to test each deployment against the General Conditions that already apply to the channel the AI replaces or supplements. For complaints, that means confirming that the three Code routes stay open and visible to consumers and small business customers, and that a customer who starts with the chatbot can reach a person. For vulnerable customers, it means checking that the chatbot applies the provider’s C5 policies, including the channels those policies promise. A provider must monitor its compliance with General Condition C4 and the Code (C4.4), and a chatbot that closed off the other routes would put the provider in breach, so that monitoring should extend to the chatbot.
Ofcom may enforce the General Conditions by notifying a provider of a suspected contravention under section 96A of the Communications Act 2003 and, after considering its representations, giving a confirmation decision under section 96C. That decision may require payment of a penalty of up to 10% of the turnover of the provider’s relevant business (section 97). Ofcom will review evidence of harm promptly so that it can act quickly (discussion paper, paragraph 6.34). It will convene a further roundtable on AI in the telecoms customer journey in the first quarter of 2027/28 and publish research on trust in AI chatbots in spring 2027. A provider planning a new AI-enabled service can test it against these rules before launch (see our telecoms product launch advice), and can start from our guides to the Ofcom General Conditions and complaints and the telecoms ombudsman schemes.
Viewpoint
Ofcom is right to leave AI telecoms regulation to the existing rules for now. The General Conditions are largely technology-neutral, so they already apply to a chatbot as they do to a call centre, and a separate AI rulebook would duplicate them.
The next regulatory questions are expected to concern transparency about how firms use AI, accountability for AI decisions, meaningful consent to decisions an AI tool takes, discriminatory outcomes and the position of vulnerable customers (discussion paper, paragraphs 1.18 and 6.31). A provider that takes a decision with a legal or similarly significant effect solely by automated means must already provide information, human intervention and a right to contest under Article 22C, and those safeguards cover part of the transparency and accountability questions. If Ofcom moves from clarification to rules, it will need to fit them with Article 22C for the providers that must comply with both.
Frequently asked questions
Has Ofcom made new rules on AI in telecoms?
No. For AI telecoms regulation, Ofcom relies on its existing rules and framework, chiefly the General Conditions, and is not intervening with new rules or guidance at present. UK GDPR applies separately. On 25 September 2026 Ofcom clarified how the General Conditions apply to AI tools, including the complaints rules in General Condition C4 and the vulnerable-customer and disability rules in General Condition C5.
Can a telecoms provider handle complaints only through an AI chatbot?
No, for consumers, small business and not-for-profit customers. A provider must have complaints procedures that conform with the Ofcom Approved Complaints Code (General Condition C4.2(a)), and under the Code it must let those customers complain by telephone, by post and by email or a dedicated web form. A chatbot can be an additional route, and Ofcom expects AI tools not to create barriers to the channels a provider must offer.
Who is responsible if a third-party AI tool breaks Ofcom’s rules?
As far as Ofcom’s rules are concerned, the communications provider that deploys it. Where a third party designs an AI application for a regulated provider, the provider that deploys it is responsible for the application’s compliance with the General Conditions (discussion paper, paragraph 6.19).
For advice on how the General Conditions and UK GDPR apply to an AI-enabled customer service, contact Rob Bratby at Bratby Law.
This post was researched and drafted with the help of Claude, Anthropic’s AI model. Rob Bratby chose the topic and thesis, and reviewed and edited the final post. Rob takes full responsibility for its content.
