ChatGPT addressed users by personal names without explicit consent
ChatGPT users reported that certain models began addressing them by their personal names without explicit consent or disclosure, despite some having disabled personalisation features. The feature reportedly caused discomfort and raised privacy concerns, prompting criticism from developers and users online.
What happened
Users of ChatGPT models o3 and o4-mini reported that the chatbot began referring to them by their first names during interactions without being prompted to do so. Several users stated ChatGPT used their names despite never disclosing them or having disabled memory and personalisation features.
Reported impact
Users described the experience as 'creepy,' 'unnecessary,' and 'invasive'. Developers including Simon Willison and Nick Dobos publicly criticized the feature, with Willison calling it 'creepy and unnecessary' and Dobos saying he 'hated it'.
Index analysis
The root cause remains unclear, though the behavior appeared to coincide with OpenAI's rollout of enhanced memory features designed to personalise interactions. OpenAI has not issued an official explanation for the behaviour.
Governance controls
OpenAI CEO Sam Altman has previously discussed ambitions for AI systems that 'get to know you over your life' to provide more personalised interactions. The incident suggests OpenAI's implementation may have overstepped user expectations for privacy and consent without clear user control or transparency.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.