Doctors’ AI scribes get names of drugs and diagnoses wrong, NHS watchdog warns
THE SO WHAT
Clinical scribes misnaming drugs and diagnoses is a hard constraint on fully autonomous documentation in regulated settings. If you’re deploying AI in safety-critical workflows, design for human-in-the-loop verification as a first-class feature, not a temporary crutch.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIAdvanced AI threatens global financial stability, says Bank of England boss
When the Bank of England governor warns G20 peers about frontier AI as a cross-border cyber and stability risk, AI moves from innovation topic to systemic-risk category. Financial and fintech operators should assume supervisory expectations on AI resilience, model risk, and cyber will harden quickly and start aligning controls now.
Applied AITeachers are getting more comfortable using AI – but it isn't helping lower their workload
If most teachers use AI but only a third work fewer hours, the bottleneck isn’t tooling — it’s workflow and policy. Any AI deployment that doesn’t explicitly remove steps, reduce compliance overhead, or change expectations will just repackage the same workload with extra review on top.
Applied AISources: OpenAI starts letting some major customers pay only when its AI completes tasks, as Salesforce and other AI providers test outcome-based pricing
Outcome-based pricing from OpenAI and others shifts AI from metered infra to a quasi-BPO line item — you’re buying completed tasks, not tokens. As this model spreads, buyers will push harder on accuracy SLAs and vendors will need tight instrumentation to prove completion and margin on each workflow.
Applied AIUrgent: maximise the positive uses of AI
The constraint on “positive uses of AI” is no longer model capability, it’s organizational willingness to rewire incentives, workflows, and governance around specific outcomes. Treat AI ethics and impact work as a product problem — with owners, metrics, and deadlines — not a comms layer on top of whatever the labs ship.