Responsible AI Use: Privacy, Bias, Copyright and Jobs

Knowing how to use AI includes knowing when not to use it, and how to use it without putting your data, company, customers or reputation at risk. This lesson, which closes the Advanced level, covers the real risks (privacy, bias, copyright, misinformation, overreliance), practical rules for professional use and how to prepare for AI's impact on jobs.

1. Privacy and confidentiality

The most frequent risk in companies. What you type or upload is processed on third-party servers and, depending on the settings, may be retained or used to improve models.

Never paste into a tool your company hasn't approved:

Good practices: use business or company-approved versions (they usually commit not to train on your data), review your tool's privacy settings, anonymize before asking for help and, if the task doesn't need the real data, ask for the formula or procedure instead of uploading the file (lesson 7).

A prompt for anonymizing with AI's help, inside a tool your company has approved (if the text is confidential, don't anonymize it in an unapproved tool: you'd already have leaked it):

Replace every piece of data in this text that identifies a person
or company: names, Social Security or ID numbers, phone numbers,
emails, addresses, account and contract numbers. Use consistent
placeholders: [PERSON_1], [COMPANY_1], [PHONE_1]… Change nothing
else. At the end, list the placeholders you used.

Expected result: "[PERSON_1] called on October 3 to complain about contract [CONTRACT_1] with [COMPANY_1]…" plus a list of placeholders. Review it by hand before using it: it's easy for a nickname, an order number or an indirect detail to slip through (a person's unique job title at a small company also identifies them).

2. Accuracy and accountability

AI can be wrong with total confidence (lesson 2), and responsibility for the result always lies with the person who uses it. The best-known U.S. case: in 2023, in Mata v. Avianca, a federal judge in Manhattan fined two lawyers and their firm $5,000 for filing a brief that cited court cases ChatGPT had invented. Similar sanctions have followed in courts across the country.

• Verify figures, names, dates, regulations, quotes and citations against original sources.
• For health, legal or financial matters, use AI to understand the topic and prepare questions, not to replace a professional.
• The bigger the impact of a mistake, the more review it needs.

3. Bias

Models learn from text written by people, and they inherit people's biases: stereotypes about gender, race, age or class. That matters most when AI is involved in decisions about people: hiring, lending, customer service. The classic example: Amazon scrapped an experimental recruiting tool after finding it penalized resumes that included the word "women's," because it had learned from a decade of mostly male hires (Reuters, 2018).

• Don't use AI as the final decision-maker on matters that affect people.
• Check results for unfair patterns (for example, a resume filter that consistently rejects a certain group).
• Explicitly ask for objective, verifiable criteria.

AI can also help you catch bias in your own writing:

Review this job posting and flag words or requirements that could
exclude candidates without reason based on gender, age, race or
national origin, disability or family status. For each one, explain
the problem and suggest neutral wording. Posting: [paste the text]

Expected result: notes such as "'digital native' can signal age bias: say 'comfortable with social media tools'" or "'must lift 50 lbs' may exclude people with disabilities if lifting isn't an essential function: describe the real requirement." Treat it as a second opinion: it can miss bias or flag non-issues.

The law points the same way: as you saw in lesson 11, U.S. anti-discrimination law covers AI hiring tools, and New York City and Illinois add specific rules for them.

4. Copyright and intellectual property

This area of law is still evolving, and lawsuits over the use of copyrighted works to train models are ongoing. What's clearer in the U.S. is the ownership of outputs. In its January 2025 report, the Copyright Office concluded that AI outputs can be protected only where a human author determined sufficient expressive elements, and that prompts alone don't meet that bar; using AI as a tool, or including AI material in a larger human work, doesn't prevent protection. In March 2025, a federal appeals court also upheld the rule that a work must have a human author to be registered (Thaler v. Perlmutter). Prudent rules:

• Don't ask for copies of protected works (books, song lyrics, paywalled articles) or imitations of characters, logos or brands.
• Review each tool's commercial-use terms, especially for images, music and video.
• If owning the result matters, add substantial human creative work: your selection, arrangement and edits are what can be protected.
• If your work will be published or sold, check your company's, client's or publisher's AI policy.

5. Misinformation, deepfakes and scams

The same technology that creates useful content can fabricate very realistic fake audio, video and images. The FTC has warned about scams that use cloned voices of family members asking for urgent money, and in 2024 the engineering firm Arup confirmed that an employee in its Hong Kong office transferred about $25 million after a video call with deepfaked executives.

Before acting on an urgent message, even if the voice or video seems real:

• I didn't act on the spot: urgency is the scammer's main tool.
• I verified through another channel: I called the number I already had, not the one in the message.
• I asked for the code word my family or team agreed on for emergencies.
• Before sharing a shocking image or video, I looked for the original source.

Our lesson on phishing and social engineering covers these tactics in depth. And never create content that impersonates real people or shows them in situations that never happened.

6. Transparency

Do you have to say you used AI? It depends on the context, but a good rule is: be transparent when the other person would care. For example, in academic work (check your institution's rules), journalism, realistic images, customer service chatbots (customers should know they're talking to AI; since 2019, California's bot disclosure law has required bots that try to drive a sale to Californians to say they're bots) or when a client pays for your creative work. Platforms and laws increasingly require labels: in the EU, since August 2, 2026, people must be told when they're interacting with AI and deepfakes must be labeled. And never overstate what your AI does: that's exactly what the FTC's "Operation AI Comply" targeted.

7. Overreliance and critical thinking

Handing everything to AI has a hidden cost: you stop practicing skills you need (writing, reasoning, deciding). Try your own approach first on learning tasks, ask for the why and not just the answer, and when a reply simply confirms what you already thought, ask for the counterargument.

AI and jobs: how to prepare

AI is automating tasks more than whole occupations. In almost any job, part of the work (drafting, summarizing, searching, processing repetitive data) can be done much faster, while other parts gain value: judgment, relationships, accountability, creativity applied to context and deep knowledge of the business.

Concrete strategies:

1. This monthBecome the person on your team who uses AI bestDocument your prompts and use cases and share them.
2. This quarterBuild complementary skillsData analysis (our data analytics course), communication, negotiation, project management.
3. AlwaysGo deeper in your fieldAI is far more useful (and less dangerous) in the hands of someone who can tell when it's wrong.
4. Every week · 1 hourKeep learningTry new features on your real work.

Your personal AI policy (template)

MY RULES FOR USING AI

1. Use: drafts, summaries, ideas, learning, formulas and code.
2. I don't upload: personal data, passwords or confidential
   information (except in tools my company has approved).
3. I anonymize before asking for help with real cases.
4. I verify: numbers, names, regulations, quotes and sources, always.
5. I review everything I send or publish; the responsibility is mine.
6. I don't delegate decisions that affect people.
7. I respect copyright, trademarks and other people's likeness.
8. I'm transparent when the other person would care.
9. I verify any urgent request for money or data through another channel.
10. I review these rules every 6 months.

Summary

• Privacy: don't paste personal data, credentials or strategic information into unapproved tools; anonymize or ask for the procedure.
• Accuracy: the responsibility is yours; the bigger the impact of an error, the more review it needs.
• Bias: don't let AI decide about people; U.S. law already covers AI hiring tools.
• Copyright: in the U.S., prompts alone don't make you an author; add real human creative work and respect others' rights.
• Scams with cloned voices and video: when money and urgency appear, always verify through another channel.
• Jobs: AI automates tasks; people who combine judgment, domain expertise and good AI use gain value.

You've finished the Advanced level! You've gone from understanding what AI is to building assistants, automating processes and planning a responsible rollout in a company. Now the Level 4: Expert section begins, where you'll build solutions businesses pay for.

Next lesson: RAG: How to Make AI Answer from Your Company's Documents. You'll learn what RAG and embeddings are, build a document search in Python and see how to prevent made-up answers.