16/04/2026
The rollout of 'AI transcription' in social work highlights a deeper problem with treating these systems as neutral helpers. A February 2026 Ada Lovelace Institute report found that AI tools meant to turn conversations into clean records were instead reshaping them—sometimes inserting serious inaccuracies like suicidal ideation, distorting children’s words, or generating outright nonsense.
What’s striking isn’t just the errors, but the way the AI shifts responsibility onto practitioners and specialists: the system produces something that looks finished and authoritative, but they’re still the ones accountable for checking every line and catching what doesn’t belong. In practice, AI transcription stops being a passive record of what was said and becomes an active, unreliable interpretation—one that can quietly alter meaning in a context where precision really matters.
A February 2026 Ada Lovelace Institute report on AI transcription tools in UK social care found that social workers were catching fabricated and mangled details in draft records, including false references to suicidal ideation, invented wording in children's accounts, and blocks of outright gibberis...
