UK regulator is looking at whether AI companion apps create harmful dependency
Key Points
- Ofcom will examine AI companion apps over concerns they create harmful dependencies and unhealthy user relationships.
- The work complements Ofcom research on trust and AI chatbots within its media literacy programme.
- Ofcom said too much trust in AI can make people believe harmful or misleading outputs.
- The regulator reported that 54% of UK adults now use AI tools, up from 31% in 2024.
- The work sits within Ofcom's wider online safety remit, including new duties under recent legislation.
Ofcom will examine the implications of AI companion apps for online safety, citing concerns the products could encourage unhealthy relationships between chatbots and users and create harmful dependencies.
The regulator set out the work in its strategic approach to AI for 2026/27, published on Thursday (4 June).
Ofcom said it recognises there are many concerns about the potential of AI companions to encourage unhealthy relationships between users and chatbots, and to create harmful dependencies.
The regulator said the work will complement its research on trust and AI chatbots, which forms part of its media literacy programme.
Data from Ofcom shows that AI adoption has risen sharply among the public. The regulator reported that over half of adults, at 54%, now say they use AI tools such as ChatGPT, Copilot or Gemini, up from 31% who said the same in 2024.
The regulator said little is currently known about how people perceive, talk about and place trust in AI chatbots, and what this means for the media literacy skills they use when engaging with them.
Ofcom said its trust and AI chatbots research will focus directly on how language shapes public use, understanding and trust in AI. The regulator said this work will build on a scarce research base.
Trust and anthropomorphisation
Ofcom said its analysis on AI and media literacy has found that users, companies and institutions often have a tendency to anthropomorphise AI chatbots, meaning they describe or treat the systems as though they have humanlike qualities.
The regulator said the ways in which language and presentation influence people’s trust in one another, trust in information and trust in automation are under researched, yet fundamental to understanding how generative AI is used.
Ofcom said examining usage, attitudes and the role of anthropomorphisation will produce insights that can support best practice in product design. The regulator said the work will improve understanding of where and how harms may develop.
Trust as a safety issue
Ofcom said the right level of trust in AI matters for consumer safety. The regulator warned that too much trust can make people more likely to believe harmful and misleading outputs, while not enough trust could make people reluctant to use certain services and feel isolated.
The regulator said AI has affected consumer safety across the sectors it regulates, creating and enhancing a number of risks. Ofcom said these include exposure to harmful deepfakes, more sophisticated fraud and scams, and the spread of misinformation and disinformation. The regulator said it aims to mitigate these risks through its policy and research work.
Ofcom said the AI companions work sits within its wider online safety programme.
The regulator said it will continue to act within its powers to address harms exacerbated by AI, and will prepare for potential new duties following the Crime and Policing Act and the Schools and Children’s Wellbeing Act.
