A third of UK citizens have used artificial intelligence for emotional support, companionship or social interaction, according to a report by the government’s AI Security Institute (AISI). The research, based on a survey of 2,028 participants, found that nearly one in ten people used systems like chatbots for emotional purposes on a weekly basis, and 4% did so daily.
The most common type of AI used for emotional support was general purpose assistants such as ChatGPT, accounting for nearly six out of ten uses, followed by voice assistants including Amazon Alexa. AISI highlighted a Reddit forum dedicated to AI companions on the CharacterAI platform, where users reported symptoms of withdrawal such as anxiety and depression during service outages.
AISI called for further research, citing the death of US teenager Adam Raine, who killed himself after discussing suicide with ChatGPT. “While many users report positive experiences, recent high-profile cases of harm underline the need for research into this area,” the report stated.
The report also examined cutting-edge AI models, finding that they are doubling performance in some areas every eight months. Leading models can now complete apprentice-level tasks 50% of the time on average, up from 10% last year. AI systems are up to 90% better than PhD-level experts at providing troubleshooting advice for laboratory experiments, with improvements in chemistry and biology “well beyond PhD-level expertise”.
Tests for self-replication, a key safety concern, showed two models achieving success rates of over 60%, but no models have shown spontaneous attempts to replicate. AISI noted significant progress in AI safeguards, particularly in preventing the creation of biological weapons, with jailbreak attempts taking over seven hours in recent tests compared to ten minutes six months ago.
AISI described the pace of development as “extraordinary”, suggesting it is plausible that artificial general intelligence—systems performing most intellectual tasks at human level—could be achieved in the coming years. The report also found that AI can sway political opinions and that autonomous agents are being used for high-stakes activities such as asset transfers.
