Companies using customer-facing artificial intelligence systems in the European Union face new transparency duties from 2 August as research indicates that almost half of UK consumers already question whether messages they receive are genuine.
Article 50 of the EU AI Act requires people to be informed when they are interacting directly with an AI system unless the automated nature of the exchange is obvious from the circumstances. The provisions also cover certain forms of synthetic or manipulated content.
UK companies can fall within the rules when their systems or services are offered in the EU, bringing exporters, software providers, digital platforms, and organisations serving European customers within scope.
Research commissioned by Exclaimer found that 46% of UK adults had questioned whether a message was genuine or legitimate, while 43% had questioned the authenticity of a wider communication.
The survey covered 1,000 adults and was originally published in June. Its findings provide context for the approaching Article 50 deadline rather than representing a new round of fieldwork.
More than half of respondents, at 53%, said the platform through which a message arrived affected how much they trusted it. Some 47% said the channel influenced whether the communication felt authentic.
Ed Bodey, general counsel at Exclaimer, said: “While the EU AI Act will require businesses to disclose AI system usage in customer conversations, our research shows that people are already questioning the authenticity of messages. Regulation is simply catching up with instinct. Nearly half of UK adults already question what lands in their inbox, and people are conscious of the giveaways in emails, whether it’s a correct company domain or a strong signature. Businesses must continue to prioritise trust in communications regardless of changes in regulation, because even the smallest detail can make or break a customer relationship.”
Half of UK respondents used AI in some aspect of their communications. Grammar and spelling improvement was selected by 22%, while the same proportion used the technology to make writing sound more professional.
Other uses involved interpersonal judgement. Some 13% used AI to soften difficult messages, 11% used it to sound more confident, and 7% used it to avoid awkward conversations.
Adoption varied significantly by age. AI was used in communications by 76% of people aged between 25 and 34, compared with 26% of those aged 65 or over. Older respondents were also more likely to question authenticity.
The transparency requirement creates a practical design problem. A disclosure must be clear enough to inform the customer without overwhelming every interaction with repetitive notices.
Companies will need to determine when a person is communicating with an automated system, when a human remains meaningfully involved, and how that distinction should be explained across websites, apps, email, messaging platforms, and telephone services.
Disclosure alone will not establish trust. Customers may know that a chatbot is automated but still receive inaccurate information, become trapped in a support loop, or be unable to reach a person.
Effective governance must therefore include escalation routes, monitoring, record keeping, and clear responsibility for decisions made with AI assistance. Systems should also be tested for accessibility and for their ability to recognise situations requiring human intervention.
Email remains important where communication has legal, employment, financial, or evidential significance. The survey found that 69% preferred email for a formal complaint, 58% for an employer update involving benefits or policy changes, and 44% for a job application or work opportunity.
Consumers also look for established markers of legitimacy. A professional company email address was important to 53%, full contact details to 51%, and a clear sender name to 33%.
Those features gain value as generative AI enables fraudulent messages to use more polished language and personal detail. Security providers are already developing controls for AI-assisted email attacks, while companies must also reduce the risk created by their own automated communications.
Implementation will require coordination between legal, customer experience, information security, marketing, and technology teams. An organisation may comply within one chatbot while overlooking AI-generated replies, virtual assistants embedded by suppliers, or automated systems operating through third-party platforms.
Article 50 establishes a minimum transparency requirement. Customer confidence will continue to depend on whether communications are accurate, secure, appropriately branded, recoverable, and supported by a human route when automation cannot resolve the issue.
You must be logged in to post a comment.