FCA authorised · FRN 724952 0117 325 0027Quote & buy →
Aipedia: Social workers
Predictive risk tools can flag families for attention, but a score is not a decision and social workers must apply their own judgement to it. Biased data can wrongly label families, with serious consequences for them.
Part of: AI in social work
In short
Some children's services use or consider algorithmic tools that combine data from different sources to score risk or prioritise cases. Social Work England has warned that biased data can wrongly label people "high risk" or "low priority", and that AI cannot replace professional judgement. Since 5 February 2026, UK GDPR Articles 22A to 22D restrict significant decisions based solely on automated processing of special category data and require safeguards such as human intervention and a right to contest. The social worker's job is to treat the score as one piece of information, test it, and record their own reasoning.
Predictive or risk-scoring tools draw on data held by a council and its partners, such as previous referrals, school attendance or housing records, and produce a score or flag. The output may be used to prioritise contacts at the front door, to suggest which families might benefit from early help, or to highlight cases for review. The design, data and thresholds are chosen by the employer and the supplier, not by the individual social worker.
SWE's research lists bias, discrimination, inaccurate outputs and unclear data handling among the risks of AI in social work, and found that accountability between social worker, line manager and software maker is often unclear.
In a GOV.UK blog in March 2026, SWE's executive director warned that AI tools trained on biased data can wrongly label people "high risk" or "low priority", and that AI cannot replace professional judgement. In children's services both errors matter. A family wrongly scored high risk may face intrusive enquiries and lasting stigma. A family wrongly scored low priority may not get help that would have prevented harm.
Data held about families reflects past contact with services, which is not spread evenly across communities. A model built on that data can reproduce the pattern and present it as objective. The score does not show what it could not see, such as a family's strengths or recent change.
The main duties that apply are:
Meaningful human review is what keeps a decision from being solely automated. Rubber-stamping a score is unlikely to be enough.
Illustrative scenario: a referral comes in about a family the tool scores as low priority because previous contacts were closed with no further action. The duty worker, with a heavy queue, accepts the rating and sends a letter. The referral contained new information that would have changed the picture if read in full. Harm follows, and a review asks why the referral was not looked at properly and what weight the tool was given.
The reverse case, a family escalated because of a high score with little else to support it, can lead to a complaint, a data protection challenge and loss of trust in the service.
The tool and the decision to use it belong to the employer, and claims about service decisions are usually handled through the employer's arrangements. Data protection complaints and any regulatory action against the council are its matter too; see whether fines are covered by PI. The individual exposure is professional: how you used your judgement can be examined in a review or, in a serious case, a fitness-to-practise process; see AI errors and fitness to practise.
You can reduce the risk with these steps:
Related: AI in safeguarding referrals and MASH triage and AI in adult eligibility decisions.
Apex arranges PI for independent social workers and consultancies advising children's services. A broker can review how your cover treats advice on or use of risk tools; see PI for social workers and children's services and the AI in social work hub.
Not on its own where the decision is significant and based solely on automated processing of special category data. UK GDPR Articles 22A to 22D, in force from 5 February 2026, require safeguards including human intervention and the right to contest. A social worker should review the case and make the judgement.
Record your own assessment, discuss it in supervision and raise it through your employer's processes. Standards 6.1 to 6.5 cover raising concerns about unsafe practice. SWE has warned that biased data can wrongly label people high risk or low priority.
Claims about service decisions are usually handled through the employer. Professionally, you are expected to apply your own judgement under standard 3.1, so following a score without looking at the evidence could be questioned in a review or regulatory process.
Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.
Start a PI quote Request a callbackApex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.
