top of page
IICE AI Code of Practice
Artificial intelligence has moved from the margins of professional life to the centre of it. Canine professionals across our sector, including trainers, behaviourists, walkers, boarders, groomers, hydrotherapists, clinical massage therapists, writers and educators, are now using AI tools daily and often without clear guidance on what counts as ethical practice. The pace of adoption has outrun the pace of professional reflection.
This Code of Practice exists to close that gap.
It is written for ethical, force-free and fear-free canine professionals who care about welfare-first practice and who want to use AI in ways that reinforce, rather than undermine, the principles their work is built on. It sets out eight core principles that should guide AI use in canine practice, alongside practical guidance for everyday situations.
This Code does not tell practitioners which AI tools to use. That choice belongs to the individual. What it does establish is what ethical AI use looks like, regardless of which tool is in front of you.
The International Institute for Canine Ethics (IICE) is committed to setting the gold standard for ethical practice in our sector. This Code is offered in that spirit and is intended to support every member of the canine professional community we serve.
Purpose
This Code establishes the ethical standards that the International Institute for Canine Ethics expects of its members in their use of artificial intelligence in professional practice. It is intended to protect dogs, clients, practitioners and the wider profession from the harms that poor AI use can cause, while supporting the genuine benefits that thoughtful AI use can offer.
Who this Code applies to
This Code applies to all IICE members across the full range of professional roles our membership covers, including:
Dog trainers and canine behaviourists
Dog walkers, home boarders and canine carers
Dog groomers
Canine hydrotherapists, masseurs and clinical bodyworkers
Dog writers, course creators and teachers
Any other canine professional whose work intersects with these areas
Different roles will encounter AI in different ways. The principles in this Code are written to apply across the full range of roles, with practical examples drawn from across the sector.
What this Code does not do
This Code does not list approved or prohibited AI tools. The AI landscape changes too quickly for that kind of list to remain useful. It also does not tell practitioners how to run their businesses or which technology to choose. The principles are tool-agnostic and durable.
A living document
This Code is published as Version 1.0. AI is evolving quickly, and so must our collective response to it. Future versions will reflect changes in technology, regulation, sector practice and the lived experience of IICE members. Feedback from practitioners is actively welcomed and will inform future revisions.
This Code does not stand alone. It sits within the wider professional framework that IICE members already operate within.
The IICE Ethics Policy
The IICE Ethics Policy is the foundational standard for ethical practice in canine professional work. It sets out the core values and obligations that all IICE members commit to. The principles in this Code are extensions of, not departures from, that foundational policy. Where this Code addresses AI specifically, the Ethics Policy addresses ethical practice as a whole.
The IICE Ethical Scholarship and Academic Integrity Standard
The IICE Ethical Scholarship and Academic Integrity Standard governs AI use in academic and learning contexts, including how learners may and may not use AI in their studies and assessments. Where the Academic Integrity Standard addresses AI use in study, this Code addresses AI use in professional practice. The two are complementary and share the same ethical spine.
The Animal Behaviour and Training Council (ABTC)
IICE members working towards or maintaining ABTC registration must continue to meet the standards set by the ABTC in their domain. Where ABTC standards apply, they take precedence within their scope. This Code complements ABTC standards by addressing AI use specifically, which the ABTC framework does not yet cover in detail.
The IICE Concerns Policy
Where a member observes a breach of this Code that they feel unable to address directly, or that involves welfare risk, the IICE Concerns Policy provides the mechanism for raising concerns formally. The Concerns Policy is available at www.canineethics.org/concerns.
UK and international law
Members must always comply with the law in the jurisdictions they practise in. This includes data protection law (UK GDPR for UK practitioners, and equivalent frameworks elsewhere), advertising and consumer protection law, and any sector-specific regulation that applies to their work. This Code sets professional expectations beyond legal minimums.
Part One - The Eight Principles
Why this matters
Behavioural and welfare judgement is the professional responsibility of the practitioner. It requires direct observation, contextual understanding and the kind of integrated thinking that AI tools cannot perform. Behaviour is the visible symptom; the root cause requires sentient, situated assessment. A dog displaying resource guarding may be in pain, in fear, in conflict or responding to handler error, and only a human professional in direct contact with the dog can distinguish between these in a specific case.
AI tools can support professional thinking. They cannot replace it.
In practice
Behavioural assessments are written by the practitioner who conducted the assessment.
Welfare judgements about an individual dog are made by the practitioner with direct contact with that dog.
AI may be used to help structure thinking, organise notes or check terminology. It must not generate the clinical conclusion itself.
Case in point
Good practice in action
Using AI to summarise your own case notes for your own review.
Using AI to suggest questions you might ask during your own consultation.
Using AI to check the spelling of a behavioural term in your draft.
Common failure modes
Submitting an AI-generated behaviour plan as the practitioner's own clinical work.
Using AI to "diagnose" a dog from a written description without direct assessment.
Relying on AI summaries as a substitute for case-specific observation.
Why this matters
Some professional decisions carry consequences that cannot be undone. A dog being bitten by another dog. A child being bitten by a dog. A dog being surrendered or euthanised because their behaviour was misread. In these high-risk areas, AI-generated advice introduces unacceptable variability and risk, regardless of how confidently it is presented.
AI should not be used to generate or support
Aggression cases or bite history assessment.
Welfare risk assessments.
Veterinary referral decisions.
Cases involving children, vulnerable adults or contested family situations.
Court reports or legal evidence.
Cases requiring multi-agency liaison.
Decisions about whether a dog should be rehomed, returned to rescue or, in the gravest cases, considered for euthanasia.
Case in point
Good practice in action
Carrying out high-risk assessments through direct, in-person professional practice.
Documenting your reasoning and evidence in your own professional voice.
Seeking peer consultation or supervision for high-stakes decisions, rather than seeking AI input.
Common failure modes
Asking AI to draft an aggression management plan for a specific client case.
Using AI to determine whether a dog needs veterinary referral.
Producing court report content through AI generation.
Using AI "behaviour analysis" tools that claim to assess dogs from video.
Why this matters
AI tools are trained on large bodies of internet text. That text includes peer-reviewed welfare science alongside outdated dominance theory, aversive training advice and harmful misinformation. AI tools cannot reliably distinguish between them. The output may sound confident and authoritative while citing or paraphrasing methods that compromise welfare.
This is the area of greatest day-to-day risk for ethical canine professionals. Force-free and fear-free practitioners cannot assume that AI output reflects their professional values. Every output relating to behaviour, training or welfare must be evaluated against the current scientific consensus on welfare-first methods.
In practice
Treat AI output as a draft requiring verification, not a finished answer.
Cross-check claims against trusted, peer-reviewed sources.
Be especially alert to language that subtly normalises aversive methods, for example "correction", "dominance", "pack leadership" and "alpha".
Never share AI-generated behaviour or training advice with clients, audiences or social media without verifying its alignment with welfare science.
Case in point
Good practice in action
Reviewing every AI output for outdated training language before use.
Cross-referencing AI claims against your professional reading and CPD.
Maintaining a personal list of trusted sources you check AI output against.
Common failure modes
Publishing AI-generated training advice without checking its methodological basis.
Forwarding AI output to clients as if it were the practitioner's own science-backed guidance.
Trusting the AI tone of authority as a proxy for scientific accuracy.
Why this matters
Authorship matters. Clients, colleagues and the wider profession have a legitimate interest in knowing whose thinking they are engaging with. When AI is used substantively in producing professional content, such as assessments, articles, client communications, training plans or course material, the practitioner must be honest about that involvement, both with themselves and, where appropriate, with their audience.
This principle aligns with the IICE Ethical Scholarship and Academic Integrity Standard, which addresses authorship in academic contexts. The same principle applies in professional contexts: your work should represent your thinking.
In practice
Use AI to support your thinking, not to substitute for it.
Where AI has substantially shaped content, ensure the final output reflects your own professional understanding and judgement.
Be especially careful with reflective writing, case studies, professional articles and CPD records. These should represent genuine human reflection
Case in point
Good practice in action
Doing your own reflective writing, even when it is harder than asking AI.
Substantively rewriting AI-drafted content so the final voice is yours.
Adding a brief note where AI was a substantive part of the process, for transparency with yourself and others.
Common failure modes
Submitting AI-generated reflective writing as personal reflection.
Publishing AI-drafted articles under your own name without substantive editorial reworking.
Using AI to fabricate case histories or professional anecdotes.
Asking AI to "write in my voice" so that the appearance of authorship is preserved while the substance is not.
Why this matters
Clients have a right to understand what they are paying for. A behaviour assessment they believed was personally written for their dog is materially different from one substantially produced by AI. Transparency does not require listing every tool used to draft an email, but it does require that clients are not misled about the human professional judgement they are receiving.
This principle is also about trust. The canine profession is built on relationships, between practitioner and client, practitioner and dog, and between practitioners and the wider community. AI use that erodes that trust ultimately damages the profession.
In practice
Be open with clients about how AI may be involved in their materials.
Avoid using AI in ways that would surprise or disappoint your clients if disclosed.
Where appropriate, include a brief AI statement in client agreements or service descriptions.
Case in point
Good practice in action
Adding a short AI use statement to your terms of service or course welcome.
Answering honestly if a client asks whether AI was involved in their materials.
Choosing to handle sensitive client communication personally rather than via AI drafting.
Common failure modes
Generating personalised reports through AI while implying bespoke human work.
Hiding AI involvement when clients ask directly.
Using AI to imitate personal communication in ways that mislead the client.
Why this matters
Most AI tools process data in ways that may train future models, store inputs, or share information with third parties. Entering client names, dog details, behavioural histories, case notes, photographs or video footage into AI tools without appropriate consent can breach data protection law and professional confidentiality alike.
Practitioners working in the UK must comply with UK GDPR. Practitioners elsewhere must comply with their local data protection framework. In all jurisdictions, professional confidentiality is a foundational duty.
In practice
Anonymise case material before entering it into AI tools.
Obtain informed consent if identifiable client or dog information is to be processed by AI.
Review the privacy and data-training settings of any AI tool you use professionally.
Avoid uploading photographs or video of identifiable people, dogs or premises without explicit consent.
Maintain a clear record of which AI tools you use and what type of data each one processes.
Case in point
Good practice in action
Reading the privacy policy of any AI tool you use for professional purposes.
Turning off data-training options where the tool offers them.
Treating client information in AI tools with the same confidentiality you would apply to your case notes.
Common failure modes
Pasting a full client case history into an AI tool to generate a plan.
Uploading client videos for AI analysis without consent.
Treating "I just asked AI" as exempt from data protection responsibilities.
Forgetting that anonymisation requires more than removing names.
Why this matters
Genuine professional development is constructive, reflective and relational. It depends on engagement with peers, mentors, ongoing reading, supervised practice and critical self-examination. AI tools can support these processes by summarising reading, suggesting questions or clarifying concepts, but they cannot replace them.
A practitioner whose understanding of behaviour, welfare or learning theory is built primarily on AI output is not adequately prepared for professional practice. Certification awarded on the basis of AI-generated work creates practitioners who hold credentials but lack capability, which is a risk to dogs, to clients and to the wider profession.
In practice
Use AI to support learning, not to complete it.
Continue investing in genuine CPD: reading, mentoring, peer dialogue, supervised practice.
Reflect on what you have learned, not just on what AI has produced for you.
Treat AI summaries as a starting point for engagement, not an endpoint.
Case in point
Good practice in action
Reading source material directly, then using AI to consolidate your own notes.
Attending live CPD events and member meetings rather than relying on AI summaries.
Discussing what you have learned with peers, mentors or supervisors.
Writing your own reflections in your own words.
Common failure modes
Using AI to complete required CPD reflection or assignments.
Substituting AI-generated content for genuine ongoing study.
Relying on AI to keep up with sector developments without independent reading.
Why this matters
AI is changing rapidly. Tools that are appropriate today may not be appropriate tomorrow. Corporate decisions by AI providers, including military partnerships, changes in data handling, shifts in pricing or access, may alter the ethical landscape overnight. Ethical AI practice therefore requires continuous reflection, not a single decision to adopt or reject.
This is part of being a professional. The world changes. Practice changes. Ethics is not a settled position but a living relationship between values and circumstances.
In practice
Periodically review which AI tools you use and why.
Stay informed about changes in AI ethics, regulation and corporate practice.
Engage with peer discussion on AI use in your sector.
Be willing to change tools, methods or practices when ethical conditions change.
Case in point
Good practice in action
Setting a regular cadence (for example, six-monthly) to review your AI use.
Following developments in AI ethics through trusted news sources.
Bringing AI questions to your IICE peer community and member meetings.
Updating your practice when new information warrants it.
Common failure modes
Treating an early choice of AI tool as permanently settled.
Ignoring developments in AI ethics that affect your practice.
Defending past AI choices because they were once acceptable, rather than evaluating them on current terms.
IICE AI Principles | Quick Reference Guide
Part Two - Practical Guidance
Much of this Code addresses areas where AI use requires caution or restraint. It is equally important to be clear about where AI can be used well. Used thoughtfully, AI can lighten administrative load, support accessibility, free up time for the parts of canine practice that matter most, and help small business owners and individual practitioners do work that would otherwise be out of reach.
The following list is not exhaustive. It is offered as a generous guide to the kinds of AI use that fit comfortably within the principles of this Code, provided the underlying principles are still respected.
Marketing and communication
Drafting blog posts, social media content, newsletters and website copy (with the principles in mind).
Generating ideas for marketing campaigns, content series or course launches.
Editing and proofreading your own writing.
Drafting email responses for routine enquiries.
Producing image content for general marketing use (with the welfare-representation caveats noted in the FAQ).
Business administration
Drafting standard business documents such as terms and conditions, policies, agreements (subject to appropriate professional review).
Creating standard client communication templates.
Scheduling support, calendar management, admin organisation.
Summarising your own notes, plans or business documents.
Course and content creation
Generating draft outlines for talks, workshops or course modules.
Creating quiz questions for your own courses (with review).
Drafting slide decks from your own source material.
Translating your own materials into other languages (with review).
Learning and thinking support
Summarising your own reading and notes for your own review.
Clarifying difficult concepts as a starting point for further study.
Suggesting areas for further reading.
Structuring your own thinking before you write.
Accessibility support
AI is genuinely valuable as an accessibility tool. For practitioners with dyslexia, ADHD, executive function difficulties, visual or hearing differences, or other accessibility needs, AI can support equal participation in professional work. This Code explicitly supports such uses, which sit comfortably
within the principles set out above.
Reorganising your own writing for clarity.
Reading text aloud, or converting speech to text.
Helping with executive function tasks such as breaking down workload into manageable steps.
Supporting time management and prioritisation.
When you are uncertain whether AI use is appropriate for a particular task, work through the questions below in order. Stop at the first question where the answer requires you to pause or change your approach.
1. Does this task involve a clinical, welfare or behavioural judgement about an individual dog?
If yes, do not use AI. This work is your professional responsibility. (Principle 1)
2. Does this task fall within a high-risk domain?
Aggression, bite history, court reports, vulnerable adults, veterinary referral, multi-agency cases, decisions about rehoming or euthanasia. If yes, do not use AI. (Principle 2)
3. Does this task involve identifiable personal data about clients, dogs or third parties?
If yes, have you anonymised the data, or obtained informed consent for AI processing? If not, pause and address data protection before proceeding. (Principle 6)
4. Does this task involve behavioural, training or welfare claims that will be acted on or shared?
If yes, do you have a clear process for verifying AI output against current welfare science? If not, pause and establish your verification process before continuing. (Principle 3)
5. Will the output be presented as your own original professional work?
If yes, will the final content represent your own thinking and authorship? If not, stop and address authorship before submitting or publishing. (Principle 4)
6. Will clients or stakeholders receive material that has been substantially shaped by AI?
If yes, consider how to be transparent about this in a proportionate way. (Principle 5)
7. Is this AI use replacing genuine professional development that you should be doing yourself?
If yes, reconsider. AI cannot substitute for your own learning, supervision or peer dialogue. (Principle 7)
8. When did you last reflect on whether your AI choices remain ethically sound?
If it has been more than six months, schedule time to review. Continuous reflection is part of ethical AI practice. (Principle 8)
If you have worked through all eight questions and none has required you to pause, you may proceed with appropriate confidence. Keep the principles in mind as you work.
Part Three - Member Standards
As a member of the International Institute for Canine Ethics, I commit to upholding this Code of Practice in my professional work. Specifically, I commit to:
1. Keeping welfare and ethical practice at the centre of every AI-related decision in my work.
2. Maintaining human professional judgement on all clinical, behavioural and welfare matters.
3. Avoiding AI use in high-risk domains where direct human practice is required.
4. Critically evaluating AI output against current welfare science before acting on or sharing it.
5. Being honest with myself and others about authorship and the role of AI in my work.
6. Being transparent with clients about AI involvement in materials they receive.
7. Protecting the privacy and confidentiality of clients, dogs and third parties.
8. Continuing to develop my professional judgement and not allowing AI to replace it.
9. Reflecting regularly on whether my AI use remains aligned with my values and this Code.
10. Raising concerns through the IICE Concerns Policy if I witness breaches that I cannot resolve directly.
Professional standards depend on the willingness of members to address breaches when they encounter them. The IICE recognises that this can be uncomfortable, particularly when breaches involve peers or colleagues. The following guidance is offered to support members in doing the right thing.
Consider direct collegial conversation first
Many breaches of this Code will be the result of unfamiliarity with AI ethics rather than deliberate misuse. A respectful conversation with the colleague concerned may be enough to address the issue. Where you feel able and safe to do so, this is often the most constructive first step.
Use the IICE Concerns Policy where direct conversation is not appropriate
Where direct conversation is not appropriate, where the breach is serious, where welfare is at risk, or where the matter has not been resolved through direct discussion, the IICE Concerns Policy provides the formal mechanism for raising concerns. The Concerns Policy is available at www.canineethics.org/concerns.
Welfare always takes priority
If welfare is at risk, do not delay. Raise the concern through the appropriate channels promptly. The welfare of dogs and the safety of clients always takes precedence over professional discomfort.
Members raising concerns in good faith are protected
IICE members who raise concerns in good faith, through the proper channels, are supported by the Institute. The Concerns Policy details the protections available to members who raise concerns honestly.
Closing Remarks
This Code of Practice does not exist to make AI use harder. It exists to make AI use better, more ethical, more aligned with welfare-first values, and more sustainable for the practitioners and the profession that depend on doing this work well.
Used well, AI can lighten administrative burden, support thinking, support accessibility, and free up time for the parts of canine practice that matter most: the dogs, the clients, the relationships and the welfare outcomes. Used poorly, it can quietly erode the standards the sector has spent decades building.
The choice belongs to each practitioner. This Code is offered as a guide to making that choice well, and as a public commitment to the standards that IICE members hold themselves to.
Glossary
The terms used in this Code are defined below. Where a term has multiple meanings in wider use, the definition given here is the one that applies within this Code.
AI terms
Artificial intelligence (AI)
A broad category of computer systems that perform tasks that typically require human cognition, such as generating text, recognising images, making predictions, or producing recommendations. In this Code, "AI" refers principally to the tools canine professionals are most likely to encounter in daily work, including generative AI and AI-powered features embedded in business platforms.
Generative AI (GenAI)
AI systems that produce new content, such as written text, images, audio or video, in response to prompts. Examples include Claude, ChatGPT, Gemini and image generation tools. Most of the practical guidance in this Code applies to generative AI.
Large language model (LLM)
The underlying technology behind most generative text AI tools. LLMs are trained on very large bodies of text and produce output by predicting plausible word sequences. They do not understand the content they produce in the way a human professional does.
Prompt and prompting
The instruction or question given to an AI tool. "Prompting" refers to the practice of designing those instructions to produce more useful output. Prompting is a skill that affects the quality and usefulness of AI output.
Hallucination
When an AI tool produces output that is plausible-sounding but factually false. AI tools can hallucinate references, statistics, quotations, study findings, professional names, legal details and almost any other type of information. This Code treats verification as the responsibility of the practitioner using the tool.
Training data
The body of text, images or other content that an AI tool was developed on. AI tools may behave in ways that reflect biases or errors in their training data. Practitioners cannot usually see or audit the training data for the tools they use.
Data processing and data training
The way an AI tool handles information that is entered into it. Some AI tools use user inputs to improve future versions of the model (data training), some store inputs for a period (data processing), and some do neither. Practitioners should understand what their chosen tools do with the information they enter.
Anonymisation
The process of removing or altering identifying information from data so that individuals (human or canine) cannot reasonably be identified from it. Effective anonymisation requires more than removing names, since combinations of details can also be identifying.
Canine professional terms
Aversive methods
Training methods that rely on the application of unpleasant stimuli or the threat of them in order to suppress or modify behaviour. The scientific consensus, reflected in the IICE Ethics Policy and across the force-free professional community, is that aversive methods compromise welfare and are not appropriate in ethical practice.
Force-free and fear-free practice
Professional practice that excludes the use of force, fear, intimidation or other aversive approaches in working with dogs. The IICE supports and represents force-free and fear-free practitioners.
Welfare science
The body of peer-reviewed scientific research that informs understanding of canine welfare, behaviour, learning, emotional life and wellbeing. Ethical practice in the canine sector is grounded in welfare science rather than in tradition, opinion, or unverified popular claims.
Evidence-based practice
Professional practice that is grounded in the best available evidence, including welfare science, peer-reviewed research and the practitioner's professional experience and judgement. AI output is not evidence.
Clinical or behavioural judgement
The professional assessment of an individual dog's behaviour, welfare needs or required intervention. In this Code, clinical and behavioural judgement is always the professional responsibility of the practitioner with direct knowledge of the case.
Frequently asked questions
Yes. Accessibility uses of AI are explicitly supported. AI as a thinking aid for neurodivergent practitioners is one of its most legitimate applications. This Code is not about restricting access to assistive technology, it is about maintaining professional judgement and welfare standards. Continue to apply the principles around authorship, transparency and data protection as you would for any other AI use.
Yes, with care. Marketing and content creation are areas where AI can lighten the load significantly. The principles still apply: ensure factual accuracy, evaluate any behaviour or training claims against welfare science (Principle 3), and consider transparency where it matters (Principle 5). If you are writing under your professional name about welfare science topics, your readers reasonably expect your professional voice and judgement.
Generally yes, with caveats. Be aware that some AI-generated images of dogs misrepresent welfare-first practice, for example by depicting aversive equipment, fearful body language presented as relaxed, or training scenarios that no ethical practitioner would set up. Review images carefully before using them. Disclose AI-generated images where misrepresentation could occur, particularly in educational content.
Be very cautious. Principle 2 places behavioural assessment in the high-risk domain category. Tools claiming automated behavioural assessment may produce confident output that bears no relationship to actual welfare reality. At present, no AI tool can reliably replace direct professional observation and judgement.
bottom of page