The recent judgment of PSA v NMC and Tchampet [2026] EWHC 141 (Admin) provides a reminder of the importance of characterising misconduct correctly.

It also illustrates the substantial risk associated with the use of unchecked AI in the preparation of legal submissions (para 4)!

The nurse at the centre of the PSA appeal, had been suspended for 9 months without review. The sanction followed a finding that during a night shift in 2019, at the home of a patient (child A, who required 1:1 care), she had slept and had also acted dishonestly in recording observations.

This followed an earlier warning by Case Examiners in 2017 regarding concerns which included sleeping on duty whilst caring for a patient (para 6). In disposing of the case with a warning, the Case Examiners had been ‘persuaded by [her] assurances of new insight…and by her assurance that she would no longer work in a community (i.e. out of hospital) setting’ (9).

At the FTP hearing in March/April 2025, Ms Tchampet, admitted all factual allegations, but denied misconduct and impairment (para 29). The Panel found both misconduct and impairment (on public interest grounds).

In determining that a sanction of suspension was appropriate, the Panel concluded (referring to the sanction guidance) that the conduct was a ‘single instance of misconduct’, there was ‘no evidence of…attitudinal problems’ and ‘no evidence of repetition…since the incident’ (para 44).

In allowing the PSA appeal and replacing the order imposed with a striking off order, the Court concluded:

• The ‘passage [in the sanction guidance] about neglect of the vulnerable strongly pointed towards striking off on the facts of this case, and it was separate from…the elements of dishonesty’ (79).
• “When deciding not to strike off, the Panel acknowledged and indeed stressed many of the points which… tended to support striking off’ (86)
• ‘It was not… a single instance of misconduct. There were many aspects to the misconduct, and it extended (because of the dishonesty, which was serious…) over weeks rather than hours.’ (90)
• ‘There did appear to be an attitudinal problem because of the similarity between the 2015 misconduct…’ (91)…In respect of the written submission relied on by the registrant’s representative, the court stated (46):
• They were ‘partly generated by Artificial Intelligence (AI)…’
• ‘three authorities [were relied on], two of which did not contain the words he attributed to them and none of which supported the propositions which he said they supported’
• ‘It was not possible to refer him to a regulator because…[he]acted as an unregulated representative’.
• ‘I warned him of the potential for him to be charged with contempt of court…but I decided it was not in this case necessary to take further steps at this time.’

The judgment is certainly well worth a read!