Is AI ethics really an individual professional concern, or a job for the companies building the models?

Both, but they are different questions. Training data, model architecture and algorithmic fairness sit with developers and platform owners. Which tasks you use AI for, what you disclose, what you do when output might be wrong, and what data you paste in are decisions you make yourself, several times a week, with real consequences.

If an AI tool produces an error in my work, who is responsible?

You are. Using AI does not create a layer of accountability sitting underneath your professional obligations; it adds a production tool inside them. A lawyer who files a brief with fabricated citations, an analyst who presents a miscalculated figure, and a clinician who acts on an unreviewed recommendation are each responsible for the output they signed off on.

When should I disclose that I used AI to produce something?

The workable test is whether the person receiving the work would reasonably want to know. A drafted client email usually does not need it. An analysis driving a significant business decision usually does. If you work in a regulated profession, check whether your regulatory body has already issued guidance. When you are unsure, disclose — the cost of unnecessary transparency is close to zero, and the cost of undisclosed involvement that later surfaces is not.

What should I check before using AI in a decision that affects a person?

Ask whether the tool has been evaluated for fairness against the groups the decision affects. Recruitment screening, performance assessment, lending and service allocation all fall into this category. If you cannot answer that question about the tool, that is itself a reason to investigate before proceeding rather than a reason to proceed carefully.

Can I put confidential client or company data into a general-purpose AI tool?

Not without checking first, and this is one of the most commonly overlooked risks because nothing visibly goes wrong. Consumer interfaces generally process your input through the provider systems. Before entering anything sensitive, confirm which data processing agreements apply, whether your organisation has approved that tool for that category of data, and what the provider retains or trains on.

AI Ethics at Work: What Every Professi…

Why AI ethics matters for individual professionals

AI ethics is often discussed as if it is only a concern for the companies building AI systems. Decisions about training data, model architecture, and algorithmic fairness are genuinely matters for developers and platform owners.

But professionals using AI at work face their own set of ethical decisions every day. Which tasks is it appropriate to use AI for? What do you tell clients or colleagues about AI involvement? What do you do when AI output might be wrong? How do you use AI in ways that do not create unfair outcomes for the people affected?

These are not abstract questions. They come up in real work situations and the decisions made there have consequences.

Accuracy and the professional responsibility problem

AI models produce outputs that are confident and fluent but not always accurate. In professional contexts, the responsibility for the accuracy of work you submit does not transfer to the AI tool you used to produce it.

A lawyer who submits a brief with fabricated case citations, produced by an AI that the lawyer did not verify, is professionally responsible for those citations. An analyst who presents figures from an AI that made a calculation error is responsible for those figures. A doctor who acts on an AI clinical recommendation without applying their own clinical judgment is responsible for that clinical decision.

Using AI does not create a new layer of accountability that sits below your professional obligations. It adds a new production tool that sits inside your existing obligations.

Warning

In any professional context where accuracy matters, establish a verification habit before adopting an AI workflow. The question is not whether AI output is usually accurate. The question is: what happens when it is not, and have I created a process that would catch that before it becomes a problem?

Transparency and disclosure

When is it appropriate to disclose that AI was involved in producing a piece of work? This varies by context and there are no universal rules yet, but the principle that guides good professional judgment is straightforward: would the person receiving this work reasonably want to know that AI was involved?

If you used AI to draft a client communication, most clients probably do not need to know. If you used AI to produce an analysis that drives a significant business decision, the person making that decision probably should know. If you work in a regulated profession where AI use has professional conduct implications, your regulatory body may already have guidance.

When in doubt, disclose. The professional risk of unnecessary transparency is close to zero. The professional risk of undisclosed AI involvement that later becomes apparent is significant.

Fairness in AI-assisted decisions

AI can embed and amplify biases present in the data it was trained on. This becomes a live professional concern when AI is used to inform decisions about people: recruitment screening, performance assessment, loan approval, content moderation, or service allocation.

Using an AI tool for a hiring task without understanding whether that tool has been tested for discriminatory outputs is not a neutral choice. The professional using the tool in a context that affects hiring outcomes carries responsibility for the outcomes.

Tip

Before using an AI tool for any task that produces a decision about a person, ask: has this tool been evaluated for fairness against the groups affected by this decision? If you cannot answer that question, that is a reason to investigate further before proceeding.

Confidentiality and data handling

Most commercial AI tools, when used through a standard consumer interface, process your inputs through the provider's systems. Entering client names, proprietary financial data, patient records, or any other confidential information into a general-purpose AI tool may violate confidentiality obligations, data protection regulations, or your organization's data policies.

This is one of the most commonly overlooked AI ethics issues at the individual professional level. The tool works well, the task is convenient, and the potential confidentiality breach is not immediately visible.

Before entering any sensitive information into an AI tool, verify: what data processing agreements apply, whether your organisation has approved this tool for this category of data, and what the tool provider's data retention and training policies are.

Intellectual property

AI-generated content can raise intellectual property questions that remain unresolved in many jurisdictions. Using AI to produce content for commercial publication, copying AI output that closely resembles a specific source, and claiming copyright over purely AI-generated work are all areas where the legal framework is still developing.

For most professional work contexts, the practical implication is to use AI for drafting and editing rather than wholesale content generation, and to maintain enough human contribution that the work is genuinely yours.

Building ethical AI habits

The professionals who handle AI ethics well do not treat it as a checklist they consult occasionally. They build habits: verify before submitting, disclose when it matters, check the sensitivity of data before entering it, and apply professional judgment to AI output rather than delegating that judgment to the tool.

The AI Fundamentals course covers the responsible use principles and professional judgment framework that applies to AI across all work contexts.

AI Ethics at Work: What Every Professional Needs to Know