AI in sensitive occupations advisor
Some professions carry confidentiality and duty-of-care obligations that ordinary AI usage can quietly violate — pasting a patient’s history, a privileged document, or a confidential source’s details into a consumer chatbot can breach the law and your professional code in one keystroke. This advisor gives profession-specific cautions and safer-practice suggestions so you can use AI without crossing those lines.
How it works
You select your occupation — doctor, lawyer, journalist, therapist, accountant, or educator — and the tool surfaces the core duty that governs your AI use (confidentiality, privilege, source protection, fiduciary duty, safeguarding). It then lists the specific cautions for that profession and concrete safer-practice alternatives. When you describe your use case, it scans for signals that confidential or identifiable information is involved and flags it as high risk. You also pick your jurisdiction, since the precise regime (HIPAA, GDPR, FERPA, bar rules) varies.
Tips and notes
- De-identify or use an approved tool. The recurring rule across every profession: don’t put identifiable confidential data into unvetted AI.
- Verify everything AI asserts. Fabricated citations and facts are a known, sanctionable failure mode in law and medicine.
- Retention terms matter. For source protection and health data, prefer tools with no-training and clear retention guarantees.
- Your regulator is the authority. This is general guidance — your professional body and employer policy are the final word.
Profession-specific risks in depth
Medical professionals
The core duty is patient confidentiality under common law and, where applicable, data protection legislation. The practical risk is using a consumer AI tool as a quick reference or documentation aid and including patient-identifiable information in the prompt. Even a seemingly harmless query — “my 47-year-old patient with hypertension and type 2 diabetes is asking about X” — can be identifiable in a small practice.
The safer path: de-identify any clinical details before they leave your system, restrict AI use to clinical decision support tools that are appropriately CE-marked or regulated (in the UK, software as a medical device falls under MHRA regulation), and use only tools with explicit data processing agreements that prohibit training on your inputs. For documentation, check whether your employer has approved specific tools for clinical correspondence — many NHS trusts and private providers are developing approved lists.
Lawyers and legal professionals
Legal professional privilege is the cornerstone. Privilege protects communications between a lawyer and client made in connection with legal advice; waiving it inadvertently — including by disclosing privileged content to a third-party AI provider — can have serious consequences. The Solicitors Regulation Authority and Bar Standards Board both expect lawyers to maintain client confidentiality and to understand the systems they use.
The fabrication problem is documented and serious: lawyers in multiple jurisdictions have been sanctioned for citing AI-generated case authorities that did not exist. The rule is absolute: never submit a citation from an AI tool without locating and reading the primary source yourself.
Journalists
Source protection is the ethical and sometimes legal cornerstone of journalism. No AI tool should ever receive information that could identify a confidential source — names, locations, communications, documents, or any detail that narrows the field. This applies both to the prompt and to documents uploaded for analysis. Investigative journalists often work with documents provided under strict source confidentiality agreements; uploading those documents to any third-party system — AI or otherwise — may constitute a breach.
For non-sensitive research tasks (fact-checking dates, generating background, drafting structure), AI tools can be useful, but the habit of treating all AI interactions as potentially logged and discoverable is the right default.
Therapists and counsellors
Therapeutic relationships involve some of the most sensitive personal disclosures imaginable. Notes, session summaries, and clinical formulations are special-category data under UK GDPR and their equivalents. The duty of confidentiality is enforced by professional bodies including the BACP, BPS, and UKCP. Using an AI tool to draft session notes or generate treatment summaries requires an appropriate data processing agreement and, in many cases, explicit client consent. Some employers are developing approved workflows; where none exists, the default should be no AI tool for client-specific content.