Responsible Use: Transparency, Integrity, Deepfakes and Critical Thinking

Knowing what AI can do and what the law says is not enough: you also need a personal sense of what is right. When should you say that AI helped you? Where is the line between assistance and cheating? How do you avoid being fooled or spreading fakes? This lesson helps you build the habits of a responsible, thoughtful user.

The previous lessons covered accuracy, bias, privacy and regulation. Now we move to behavior: the choices that no law will make for you and that define how trustworthy you are as a student, a professional and a member of the public.

Transparency: when to mention that you used AI

There is no single universal rule, but there is a useful principle: do not let people believe something false about how a piece of work was produced when it matters to them. Ask yourself whether the reader, the teacher, the client or the audience would reasonably expect to know. Consider these situations:

• Minor assistance (fixing spelling, suggesting synonyms): usually no mention needed, just as you would not mention a spellchecker.
• Substantial drafting (the AI wrote most of a text you then edited): a mention is often expected, especially in work that is evaluated, published or signed.
• Content that looks real but is generated (images, voices, "testimonials"): a clear label is generally the right thing to do, and may be legally required in certain cases.
• Interactions (a chatbot answering customers): people should know they are dealing with an AI.

When you do mention it, be simple and specific: "I used an AI assistant to generate a first outline and to check the grammar. I wrote and verified the content." Your school, employer, publisher or professional body may have a policy: follow it, and if there is none, ask.

Academic and professional integrity

Using AI is not automatically cheating, and refusing to learn it is not automatically virtuous. The question is whether your use respects the purpose of the task and the rules that apply.

In education

• Check the rules for each course or assignment. They may allow AI for brainstorming but not for drafting, or forbid it entirely for an assessment.
• Remember the purpose: an essay or exercise is a way to learn. If the AI does the thinking, you lose the learning, even when nobody notices.
• Use AI as a tutor: ask it to quiz you, explain your mistakes, or suggest how to improve your own draft.
• Submitting AI-generated work as your own where it is not allowed is usually considered misconduct, with consequences.
• Do not rely on "AI detectors" as proof either way; they can be wrong and can unfairly accuse people. Keep drafts and notes that show your process.

At work

• Follow your organization's policy on approved tools and acceptable uses.
• You are accountable for what you deliver. "The AI wrote it" is not an excuse for errors, plagiarism or leaks.
• Be honest about your process with colleagues and clients when it matters to quality, confidentiality or contracts.
• Do not pass off generated expertise as professional qualifications. For advice in regulated fields (medicine, law, finance), the responsible professional remains a human with the right credentials.

Deepfakes and disinformation

Generative AI makes it cheap and quick to create realistic images, voices, videos and texts. A deepfake is a synthetic or manipulated media that shows someone saying or doing something they did not. The same technology can be harmless (a parody clearly labeled as such, a film effect) or harmful (fraud, harassment, fake news, impersonation).

Risks

• Fraud and impersonation: a cloned voice asking a relative for urgent money; a fake message from a "manager."
• Reputation and harassment: fake images of real people, especially intimate or humiliating ones, which can be illegal and deeply harmful.
• Disinformation at scale: large volumes of convincing but false texts and images, especially around elections and crises.
• The "liar's dividend": when everything can be faked, real evidence can be dismissed as fake.

What you can do

• Never create realistic fake media of a real person without their clear consent, and do not use AI to deceive, harass or defraud.
• Pause before sharing. Surprising, emotional content that confirms what you already believe deserves extra scrutiny.
• Verify the source. Who published it first? Do reliable outlets report the same? Can you find the original?
• Look at the details, but do not rely on them. Odd hands, inconsistent shadows, strange text or audio glitches can be clues, but good fakes may have none, and real media can look odd. Provenance matters more than pixel-peeping.
• Use a second channel for urgent requests. If a voice or message asks for money or sensitive information, call the person back on a number you already know. Some families and teams agree on a code word.
• Label what you generate when it could be mistaken for real.

Environmental and social impact

Training and running AI models consumes electricity, water for cooling data centers, and hardware resources. Without quoting figures that change quickly, it is fair to say that the impact is real, that it grows with usage, and that bigger models and heavier tasks (such as generating video) generally require more resources than a short text request. Providers are working on efficiency, but the overall effect depends on how much AI is used.

What you can do individually is modest but meaningful: use AI where it adds real value, not out of habit; keep prompts focused instead of regenerating endlessly; choose lighter tools or features for simple tasks; and do not automate pointless work. There are also social questions, such as the working conditions of people who label data, the effect on certain jobs, and who benefits from the technology. You do not need to solve these, but being aware of them is part of being a thoughtful user.

Keeping your critical thinking sharp

The biggest risk of a very helpful assistant may be a quiet one: relying on it so much that you stop thinking, remembering or questioning. A few healthy habits:

• Think first, then ask. Try the problem or draft your own outline before consulting the AI, especially when learning.
• Watch for flattery and over-agreement. Assistants often tend to agree with you. Ask for counterarguments and weaknesses.
• Keep your own voice. Edit the output until it sounds like you and says what you mean.
• Maintain core skills. Keep practicing the abilities you do not want to lose: writing, calculating, reasoning, reading sources.
• Stay curious about limits. When the tool surprises you, in a good or bad way, find out why.

Your turn: write your own position

Exercise 1. For each situation, say whether you would mention AI and how: (a) a student uses AI to explain a concept, then writes the essay alone; (b) an employee has AI draft a client proposal, then reviews and edits it; (c) a marketer generates a realistic image of a "customer" for an ad.

Exercise 2 (project). Start your "personal charter for responsible use" with five short rules, for example one on transparency, one on integrity, one on verification, one on data and one on critical thinking. You will complete it in chapter 6.

See a solution

• (a) No mention is usually needed for understanding a concept, but follow the course rules. If asked, you can say you used AI as a tutor.
• (b) Substantial drafting: follow organization policy, and be ready to say how it was produced. The employee remains accountable for accuracy and confidentiality (no client data in an unapproved tool).
• (c) The image is generated and could be mistaken for a real person: label it clearly as illustrative or AI-generated, and check the rules for advertising and testimonials in your market.

Example charter rules: "1) I mention AI when someone would reasonably expect to know. 2) I never submit AI work where it is forbidden. 3) I verify facts before I share them. 4) I do not paste personal or confidential data. 5) I think first, then ask, and I keep ownership of the final text."