BACP Ethical Framework 2026 and AI: what counsellors must do before using an AI note-taker

An AI note-taker is a tool that turns what is said in a session into text and then into a draft note that the practitioner checks. Until now, the BACP Ethical Framework did not mention AI by name. The 2026 version does. BACP says the core responsibilities section includes new guidance on the use of AI and digital technology.

This guide is for counsellors and psychotherapists who are BACP members and who use, or are thinking about using, an AI tool for session notes. It sets out what the new clause says, when it applies, and how to turn it into a checklist you can actually complete. It ends with the questions to put to any provider.

When the 2026 Ethical Framework becomes mandatory

BACP's page on preparing for the new framework says it becomes mandatory for all members at midday on Tuesday 3 November 2026. Until then, members keep using the Ethical Framework for the Counselling Professions 2018. The framework document itself states that it takes effect from 3 November 2026.

The framework applies to all members and registrants of BACP. It is meant to be the basis of all their professional work. So if you are a BACP member and you already use an AI note-taker, you have a few weeks to check your practice against the new wording.

BACP is also publishing support material. Its preparation page lists Ethics Essentials videos, audio versions of each section, a glossary and a set of FAQs on the 2026 framework. It says further e-learning and Good Practice in Action resources are being updated or developed.

What clause 2.1(e) actually says

Clause 2.1 sits under the heading about the safety of practitioners and the people they work with. Point (e) asks members to "assess the risk of any artificial intelligence tool (AI), digital tool or online platform, before using them."

Members must then be able to demonstrate five things. In plain terms, they are:

1. You are competent to use the tool or platform.
2. You understand how data are handled and stored, and any risks to confidentiality, and you have reduced those risks as far as possible.
3. You are honest and transparent with people about your use of AI and digital tools, including the benefits and the risks.
4. You have the informed consent of the people using your services before you enter any of their personal data into AI or digital tools.
5. You "have not deferred or outsourced" your decision-making to AI tools.

The clause closes with a clear line on responsibility. You remain responsible for critically evaluating AI output and for any decision you make based on it.

Note the scope. The clause covers any AI tool, digital tool or online platform. Read plainly, that wording would also take in a video call platform or a practice management system, not only an AI note-taker. The AI-specific points (consent before input, no outsourced decisions) are simply sharper for AI.

Clauses 3.1 and 3.2: privacy notices and notes

Two other clauses matter for anyone using an AI note-taker.

Clause 3.1 is about confidentiality and privacy. It asks members to understand and comply with data protection law in their country of practice. For the UK, the framework names the Data Protection Act 2018 and the UK GDPR. It also asks for a clear and accessible privacy notice. That notice must make people aware of foreseeable limits to confidentiality, including "the use of digital storage systems, platforms or tools that may monitor or collect data." An AI note-taker belongs in your privacy notice.

Clause 3.2 is about keeping appropriate records. Records must be factual, adequate and relevant for the type of service. They must be stored securely and comply with the data protection requirements of the country where the record is stored. They include all communications, notes and other information held about clients. A draft written by an AI tool, once you keep it, is part of that record. It has to meet the same standard as a note you typed yourself.

What about "never ethical"?

Many counsellors remember a Therapy Today piece from July and August 2025, "The big issue: What every therapist needs to know about AI", written by Sally Brown. It includes the line "Inputting details of real client sessions into AI is never ethical". That article is journalism in BACP's member magazine. It is not the Ethical Framework.

The 2026 framework takes a different route. It does not ban AI tools. It sets conditions: assess the risk, understand the data, be transparent, get informed consent first, and keep the decisions yours. The article's concern still deserves respect. Pasting session details into a general chatbot with no contract and no idea where the data goes would be very hard to justify under clause 2.1(e). A tool built for clinical notes, under a contract, with clear data handling and the client's informed consent, is a different situation. You still have to check it yourself.

A practical checklist before you use an AI note-taker

Here is clause 2.1(e) turned into tasks. Keep a written copy of each step. The clause says you must be able to demonstrate these points, and a dated document is the simplest way to do that.

1. Assess the risk of the tool

• Write down what the tool does, what data it receives (audio, transcript, names, health details) and what it produces.
• List the risks: a breach, data kept longer than needed, a wrong or invented detail in a note, a client feeling observed.
• For each risk, note what reduces it and who is responsible.
• Consider a data protection impact assessment (DPIA). The ICO lists innovative technology, including AI, among the types of processing that can require one. It also says that if you are in any doubt, it recommends doing a DPIA, and that you should document your reasons if you decide not to.

2. Be competent to use it

• Read the provider's documentation and try the tool on a mock session before using it with a client.
• Learn how to pause it, delete a draft and correct a note.
• Know its limits. Transcription can mishear names, accents and quiet speech.
• Bring the topic to supervision.

3. Understand how data are handled and stored

• Where is the data hosted, and where does the AI processing run?
• Is the audio kept, and if so for how long?
• Is client data used to train AI models?
• Are names and identifying details removed or replaced before AI processing?
• Is data encrypted in transit and at rest?
• Can you get a data processing agreement and a list of subprocessors?
• How do you delete a client's data when your retention period ends?

If a provider cannot answer these questions in writing, that is your answer.

4. Be honest and transparent

• Update your privacy notice to name the type of tool, what it does and where data goes (clause 3.1).
• Add a short paragraph to your counselling contract.
• Explain the benefits and the risks in plain words, not only the benefits.

5. Get informed consent before any personal data goes in

• Ask before the first session where you would use the tool, not halfway through it.
• Make it a real choice. Offer to work without the tool, with no change to the service.
• Tell clients they can withdraw consent at any time, including for a single session.
• Note the consent and its date in the client's file.

For wording you can adapt, see our guide to AI note-taking and client consent in the UK.

6. Keep the decisions yours

• Read every draft in full before you save it. Correct it, cut it, or reject it.
• Do not let a summary replace your own judgement about risk, safeguarding or the direction of the work.
• Remember that the note you sign is your note, whatever produced the first draft.

The NHS England guidance on ambient scribing products points the same way. It was written for health and care settings in England rather than for independent counsellors. Still, it asks settings to explain to patients how their information will be used before the processing takes place, giving them the chance to object. It also asks settings to ensure users review and approve any product outputs before further action. The broader risks of AI in clinical work are covered in our article on AI in clinical psychology and its risks, and confidentiality in professional confidentiality and AI.

How Delta answers each point

Delta is an AI assistant for mental health and allied health practitioners. Here is how it maps to the checklist. It does not do the checklist for you.

• Risk assessment. Delta gives you facts to put in your own risk assessment. The assessment itself stays yours to write.
• Competence. During the session, Delta transcribes what is said, then prepares a session report. You can also dictate observations right after the session or add written notes. A 14-day free trial without a card lets you test it on mock sessions first.
• How data are handled and stored. Data is hosted in France with a host certified for health data (HDS, the French health data hosting certification). AI processing, transcription included, runs on servers located in France. The client's name and identifying details are pseudonymised before AI processing. No audio file is kept. Data is never used to train models. Data is encrypted in transit and at rest. Details are on the security page.
• Transparency and consent. These depend on you. Name the tool type in your privacy notice and ask for consent before the first session with it.
• Decisions. You review and validate every report before it goes into the client's file. Delta takes into account your profession and therapeutic approach, such as person-centred, psychodynamic, CBT or systemic work. The draft is a starting point. The clinical thinking stays with you.

As with any provider, ask Delta for its data processing agreement before you start, and read it. Then complete your own risk assessment. More on how Delta fits counselling practice is on the page for therapists.

Sources

• BACP, Ethical Framework for the Counselling Professions 2026 (PDF)
• BACP, Preparing for the 2026 Ethical Framework
• BACP Therapy Today, July/August 2025, The big issue: What every therapist needs to know about AI
• NHS England, Guidance on the use of AI-enabled ambient scribing products in health and care settings
• ICO, When do we need to do a DPIA?