AI Fundamentals and Responsible Use Manual - Accessible Learning Center

First-time keyboard guide

Use this short routine before following the AI Fundamentals and Responsible Use Manual procedures. It explains exactly how the keyboard directions in this manual work.

Press a single key

Find the named key without pressing it if possible.
Press and release the key one time. Do not hold it unless the step specifically says to hold it.
Stop and listen to the full screen-reader announcement before pressing another key.

Press a keyboard combination

When a command contains a plus sign, such as Ctrl+S, hold down the first key.
While continuing to hold the first key, press and release the final key once.
Release the first key. The plus sign is not a key and should not be pressed.
Listen for the expected result described in the procedure. If nothing changes, verify that the correct application or webpage has focus and try the command once more.

Press a sequence of keys

When keys are separated by commas or the word “then,” press and release them in order instead of holding them together.
After each key, listen for a menu, tab, control, or new screen before continuing.
If the announcement does not match the step, press Escape once to close the current menu or dialog and restart that procedure.

Know which navigation key to try

Tab and Shift+Tab: move forward and backward among buttons, links, fields, and other controls.
Arrow keys: move inside menus, lists, groups of choices, text, and many screen-reader reading views.
Enter: opens the selected item or activates the focused command.
Space: activates many buttons and checks or clears checkboxes. In a text field, Space types a space.
Escape: cancels or closes the current menu, list, or dialog without intentionally accepting a change.

Verify before moving on

Listen for the window, page, or application name so you know where focus is.
Perform only one numbered step.
Compare what the screen reader announces with the “What should happen” information in the procedure.
If the result is wrong, stop. Use the procedure’s recovery directions, Escape to cancel, or Ctrl+Z to undo when that command is supported.
Continue only after you can identify the result. Save completed work with Ctrl+S whenever the application supports saving.

First complete independent workflow

This rehearsal joins existing procedures into one uninterrupted session without repeating their directions. Follow the linked procedure when a step names it, then return here for the next verification.

Safety check: Confirm the active application, document or project name, and save location before changing content.

Open AI Fundamentals and Responsible Use Manual and wait until its name, starting screen, or connected-device status is announced.
Choose a fictional, public, disposable, or instructor-provided practice item. Do not use private client, student, medical, financial, or account-recovery information.
Open 8.1 Test an AI Claim. Read its “What you will learn,” “Before you begin,” keyboard-command, and expected-result information before pressing anything.
Complete that procedure one numbered action at a time. After each key, stop and identify what changed before continuing.
State the result aloud or record it in a practice note. If it differs from “What should happen,” use that procedure’s recovery directions before advancing.
Continue directly to 9.1 Write a Basic Prompt without closing the application or device session. First confirm that focus, account, file, book, conversation, or project context is still correct.
Complete the second procedure and verify the resulting name, state, location, permission, value, or content. Do not rely only on a success sound.
Save or synchronize when the product supports it. Wait for the completed state before disconnecting a device, closing a browser, or switching accounts.
Continue to 12.1 Check a Citation. Use the exact command listed there; if the command varies by model or version, use Keyboard Help, the product’s shortcut list, or the linked official reference instead of guessing.
Introduce one harmless recoverable error, such as opening the wrong menu or moving focus away. Press Escape once, identify the active window or device area, and return to the correct step.
Review the finished practice item from its beginning. Confirm its title or filename, structure, content, destination, sharing state, and saved or synchronized status as applicable.
Close and reopen the item or return to it from the product’s main screen. The workflow is complete only when you can locate it and verify the same result independently.

Success standard: Complete all three linked tasks, explain the commands used, recover from the practice error, and locate the finished result without another person directing the next key.

Purpose: This manual explains artificial intelligence from the beginning and teaches how to use AI tools responsibly. It assumes no prior technical knowledge.

Who this is for: Students, employees, job seekers, independent learners, screen-reader and braille users, and anyone who wants to understand AI before relying on it.

Core principle: AI can be useful without being authoritative. Treat AI output as assistance that may require verification, not as automatic truth.

Responsible-use foundation: The guide reflects current responsible-AI principles from NIST and UNESCO, including risk management, transparency, fairness, privacy, human oversight, and awareness of generative-AI-specific risks.

How to Use This Manual

Read Parts 1 through 8 before using AI for important work; they establish the concepts needed to judge AI output.
Practice with low-risk tasks first, such as brainstorming, rewriting your own text, or generating study questions.
Do not paste confidential, private, regulated, or employer/client data into an AI system unless the relevant policy and service terms permit it.
Verify important factual claims using trustworthy independent sources.
Keep the human decision-maker responsible for important decisions.
Use AI to support accessibility and independence, but check outputs that describe images, documents, people, or environments.
When an AI tool cannot provide sources, treat precise factual claims with extra caution.
Keep records of significant AI use when school, workplace, legal, or organizational policy requires disclosure.

Part 1 - What Artificial Intelligence Is

Artificial intelligence is a broad term for computer systems designed to perform tasks that people associate with intelligent behavior, such as recognizing patterns, understanding language, generating content, making predictions, recommending actions, or controlling devices.

AI is not one single technology.
An AI system may be narrow and specialized, such as spam detection or speech recognition.
An AI system may generate new content, such as text, images, audio, video, or code.
An AI system may recommend or rank options without generating visible content.
AI output is produced by a system, not by an independent human expert reviewing your situation.

Important distinction: A system can produce a fluent answer without having human experience, personal awareness, or guaranteed understanding of the real-world situation.

Part 2 - AI, Machine Learning, Deep Learning, and Generative AI

Topic	What to remember
Artificial intelligence (AI)	The broad field of systems that perform tasks associated with intelligent behavior.
Machine learning (ML)	A way of building systems that learn patterns from data rather than relying only on hand-written rules.
Deep learning	Machine learning using multi-layer neural networks; widely used in speech, vision, and modern generative systems.
Generative AI	AI that creates new content such as text, images, audio, video, or code.
Large language model (LLM)	A generative model trained to work with language and related symbolic information.

Relationship: Generative AI is part of AI. Many current generative systems use deep-learning methods, but not every AI system is generative.

Part 3 - How Generative AI Produces an Answer

A generative model receives your input, combines it with the conversation/context and system instructions, and predicts a useful continuation. For language models, this involves predicting sequences of tokens.

The answer is generated at request time; it is not normally retrieved as one complete prewritten paragraph.
Generation is influenced by the prompt, prior conversation, system instructions, model design, and available tools/data.
A fluent answer can contain invented details because plausible language is not the same as verified truth.
Different prompts can lead to different answers to the same underlying question.

Mental model: Think of generative AI as a powerful pattern-and-language engine with reasoning capabilities that still needs verification, not as an infallible encyclopedia.

Part 4 - Large Language Models and Tokens

Large language models process text in smaller units called tokens. A token can be a whole word, part of a word, punctuation, or another unit.

Long conversations consume context space.
The model may pay less attention to older details when a conversation becomes very long.
Exact token counts depend on the model and tokenizer.
Giving clear, relevant context is usually more useful than giving large amounts of unrelated text.

Practical lesson: If an AI starts forgetting instructions in a long conversation, restate the key requirements or begin a clean conversation with a concise summary.

Part 5 - Training Data, Context, and Knowledge Limits

Models learn statistical patterns during training. At use time, a model may also receive current context, uploaded files, retrieved information, or web/search results if the product supports those tools.

Training data is not the same thing as a live database.
A model may not know that a fact has changed unless it has current retrieval/search access.
A model may mix accurate information with outdated or incorrect information.
An answer about current laws, prices, schedules, public figures, software versions, or news should be verified with current sources.
Uploaded documents and connected data can improve relevance but increase privacy responsibility.

Current information rule: For time-sensitive questions, ask the AI to verify with current sources when the product can browse or search.

Part 6 - What AI Is Good At

Brainstorming ideas and alternatives.
Explaining a concept at different reading levels.
Summarizing text you provide.
Rewriting your own material for clarity, tone, or structure.
Generating practice questions or study aids.
Creating outlines, checklists, templates, and first drafts.
Classifying or organizing information with human review.
Translating or simplifying material, with verification for important use.
Helping with code, formulas, and troubleshooting when results are tested.
Accessibility support such as describing visual structure or converting information into a more usable format.

Best use pattern: AI is often strongest when the user can evaluate the result or verify it independently.

Part 7 - What AI Is Not Good At

Guaranteeing that a factual statement is true.
Knowing unstated facts about a person, event, image, or organization.
Replacing professional judgment in high-stakes medical, legal, financial, employment, or safety decisions.
Determining whether an accusation is true without reliable evidence.
Knowing whether a source exists just because it can generate a plausible citation.
Understanding every local policy, law, procedure, or cultural context automatically.
Making an important decision without human accountability.

Warning sign: The more serious the consequence of being wrong, the stronger the verification and human-review requirement should be.

Part 8 - Hallucinations and Confident Errors

A hallucination is an AI-generated statement that sounds plausible but is unsupported, false, or invented. NIST's generative-AI risk profile treats confabulation and related misinformation risks as important issues to manage.

Invented references, article titles, court cases, statistics, quotes, or links.
Correct-sounding but false explanations.
Combining details from two different people, products, or events.
Making up settings, buttons, commands, or software features.
Claiming certainty where the evidence is weak.

8.1 Test an AI Claim

What you will learn: Verify a factual answer before using it.

Before you begin: Choose a low-risk factual claim.

Exactly how to do it

Ask the AI to state the claim clearly.
Ask for the source or evidence.
Open or independently locate the source.
Check that the source actually says what the AI claimed.
Check the publication/update date.
Check whether another trustworthy source agrees.
Correct your notes if the AI was wrong.

What should happen: You separate fluency from evidence.

If that does not happen: If the AI cannot provide a verifiable source, treat the claim as unverified rather than repeating it as fact.

Practice: Verify three AI-generated claims.

Part 9 - Prompting Fundamentals

A prompt is the instruction or information you give the AI. Good prompts reduce ambiguity but do not guarantee correctness.

Topic	What to remember
Goal	What do you want the AI to produce?
Context	What background is needed to understand the task?
Constraints	Length, tone, format, audience, sources, exclusions, or other requirements.
Input	The text, data, problem, or material to work on.
Verification	What should be checked, cited, or marked as uncertain?

9.1 Write a Basic Prompt

What you will learn: Give enough information without oversharing.

Before you begin: Choose a harmless task.

Exactly how to do it

State the goal in one sentence.
Add the audience or purpose.
Add important constraints.
Provide only the input needed.
Ask the AI to identify uncertainty when appropriate.

What should happen: The response is more relevant and structured.

If that does not happen: If the response misses the task, identify which prompt element was vague and revise it.

Practice: Prompt the AI to explain one familiar concept for a beginner.

Part 10 - Better Prompt Structure

A reusable prompt structure is: Task - Context - Requirements - Output Format - Verification.

Topic	What to remember
Task	Summarize this policy.
Context	I am a new employee with no technical background.
Requirements	Explain key duties and deadlines; do not invent missing information.
Output format	Use headings and a final checklist.
Verification	Quote no more than short necessary phrases and flag anything unclear.

Avoid fake authority: Telling an AI to 'act as the world's greatest expert' may change style, but it does not create credentials or guarantee accuracy.

Part 11 - Follow-Up Questions and Iteration

Ask for clarification: 'Which part of my request was ambiguous?'
Ask for alternatives: 'Give me three approaches with tradeoffs.'
Ask for compression: 'Reduce this to a five-item checklist.'
Ask for expansion: 'Explain step 3 in beginner terms.'
Ask for verification: 'Which claims here need external checking?'
Ask for assumptions: 'List the assumptions you made.'

Iteration rule: Use follow-up questions to improve usefulness, but do not confuse repeated agreement with independent verification.

Part 12 - Sources, Search, and Grounding

Grounding means tying an answer to external information such as retrieved documents, databases, tools, or web sources.

A sourced answer is stronger only if the sources are relevant and accurately represented.
A citation can be wrong, outdated, or unrelated.
Primary sources are often best for laws, policies, official product instructions, standards, and statistics.
Secondary sources can be useful for explanation and comparison.
Do not assume an AI-generated bibliography is real until you verify each item.

12.1 Check a Citation

What you will learn: Confirm that a cited source exists and supports the claim.

Before you begin: The AI supplied a source.

Exactly how to do it

Open the source through a trusted path.
Confirm title, publisher/author, and date.
Find the relevant passage.
Compare the source's wording/meaning with the AI's paraphrase.
Check whether the AI omitted an important limitation.
Record the source separately if you will rely on it.

What should happen: You know whether the citation genuinely supports the answer.

If that does not happen: If the source cannot be found, treat the citation as potentially fabricated.

Practice: Verify one article and one official source.

Part 13 - Verification and Fact-Checking

A practical verification ladder:

Topic	What to remember
Low stakes	Quick plausibility check may be enough.
Moderate stakes	Check one or two trustworthy independent sources.
High stakes	Use authoritative primary sources and qualified human review.
Irreversible action	Do not proceed from AI output alone.

13.1 Verify an Important Answer

What you will learn: Use a structured evidence process.

Before you begin: The result could affect money, health, school, work, legal rights, safety, or another person.

Exactly how to do it

Separate facts from suggestions/opinions.
List the factual claims that matter to the decision.
Identify the best authoritative source for each claim.
Check current information.
Ask a qualified human when professional judgment is needed.
Document what you verified.
Make the decision yourself or through the responsible human process.

What should happen: The AI becomes one input rather than the final authority.

If that does not happen: If reliable evidence conflicts with the AI, trust the verified evidence and correct the AI-derived work.

Practice: Practice on a harmless policy question.

Part 14 - Privacy and Sensitive Information

Do not paste passwords, security codes, private keys, or recovery codes into a general AI chat.
Do not upload confidential client, patient, student, employee, financial, legal, or proprietary data unless policy and service terms explicitly allow it.
Remove unnecessary names, account numbers, addresses, IDs, and other identifying details.
Know whether your organization provides an approved enterprise AI tool with specific data protections.
Review the service's privacy/data-control settings before using it for sensitive work.

14.1 Minimize Data Before Prompting

What you will learn: Give the AI what it needs without giving it everything.

Before you begin: Use a fictional or sanitized example.

Exactly how to do it

Identify what the task actually requires.
Remove names and direct identifiers.
Replace real account or case numbers with placeholders.
Generalize dates/locations when exact details are not necessary.
Submit the minimized version.
Check the answer before reconnecting it to real records.

What should happen: The AI can help while receiving less sensitive data.

If that does not happen: If the task cannot be done without protected data, use an approved system or do not use AI for that task.

Practice: Sanitize a fictional case description.

Part 15 - Security and Prompt Injection

Prompt injection is an attempt to make an AI system follow malicious or untrusted instructions embedded in webpages, documents, emails, tool output, or other content.

Treat instructions inside untrusted content as data, not automatically as commands.
Do not allow an AI agent to send messages, move money, delete files, or publish content without appropriate confirmation.
Be cautious when an AI says a webpage/document instructed it to ignore previous rules.
Do not expose credentials or secrets to satisfy instructions found in retrieved content.
Use least privilege: give AI tools only the permissions needed for the task.

Security principle: Untrusted content should not be allowed to silently change the user's real-world intent.

Part 16 - Bias, Fairness, and Representation

AI can reproduce or amplify patterns and biases present in data, design choices, evaluation methods, or deployment context.

Do not use AI stereotypes as evidence about a person or group.
Review hiring, admissions, discipline, lending, housing, benefits, and other consequential outputs for unfair patterns.
Test whether wording changes associated with identity characteristics change the AI's recommendation.
Include diverse perspectives when evaluating systems that affect diverse populations.
Give people meaningful human review and appeal processes when AI influences important decisions.

UNESCO principle: UNESCO's AI ethics framework emphasizes human dignity, fairness and non-discrimination, transparency, inclusiveness, and human oversight.

Part 17 - Copyright, Attribution, and Ownership

AI-generated content can raise copyright, licensing, attribution, plagiarism, and ownership questions. Rules vary by country, platform, employer, school, and intended use.

Do not assume AI output is automatically free of third-party rights.
Do not ask AI to reproduce a copyrighted work when you do not have permission.
Verify quotations and citations.
Follow employer, publisher, school, and platform disclosure rules.
Keep records of substantial human editing and creative contribution where relevant.

U.S. Copyright Office: The U.S. Copyright Office's January 2025 report states that copyright protection can apply where a human author contributes sufficient expressive elements, while merely providing prompts is not by itself enough for authorship of AI-generated output.

Not legal advice: For significant commercial or legal decisions, consult current law and qualified counsel.

Part 18 - Education and Academic Integrity

Read the assignment/school AI policy before using AI.
Use AI for tutoring, practice, brainstorming, and feedback only when permitted.
Do not submit AI-generated work as your own if the rules prohibit it.
Do not invent citations or sources.
Disclose AI assistance when required.
Keep enough of your own process to explain and defend the submitted work.

18.1 Use AI as a Study Partner

What you will learn: Learn rather than outsource the learning.

Before you begin: Choose a topic you are studying.

Exactly how to do it

Ask AI to explain the concept.
Close the answer.
Explain the concept in your own words.
Ask AI to quiz you.
Answer without copying.
Verify any disputed facts using course materials.

What should happen: AI supports recall and understanding rather than replacing your work.

If that does not happen: If you cannot explain the answer independently, continue studying before submitting anything.

Practice: Complete one five-question quiz.

Part 19 - Workplace and Professional Use

Use only approved AI tools for company/client information.
Do not paste confidential documents into personal AI accounts.
Review generated emails, reports, code, formulas, and analyses before sending or deploying them.
Verify numbers, names, dates, citations, and commitments.
Disclose AI use when policy, client contract, regulation, or professional standards require it.
Do not let AI send external communications or execute actions beyond your authority.

Accountability: The person or organization using AI remains responsible for professional output, even when the AI produced the first draft.

Part 20 - High-Stakes Decisions

High-stakes use means an error can significantly affect health, safety, legal rights, finances, employment, education, housing, insurance, public benefits, or another person's opportunities.

Use AI to organize questions or explain terms, not to replace qualified decision-makers.
Verify current law, policy, medication, diagnosis, financial data, or safety information independently.
Do not make adverse decisions about a person solely from an AI-generated profile or score.
Provide human review for consequential decisions.
Use documented standards and evidence rather than AI confidence.

Rule: The higher the consequence, the less appropriate it is to rely on unsupervised AI output.

Part 21 - Accessibility and Assistive Technology

AI can improve access by converting information into alternative forms, but accessibility output still requires verification.

Ask AI to describe complex images, charts, or layouts, then verify important details when possible.
Use AI to convert visual instructions into ordered keyboard/screen-reader steps.
Ask for plain-language or structured versions of inaccessible text.
Use AI to generate alt-text drafts, but confirm the description reflects the actual image and purpose.
Do not assume AI-generated descriptions identify people, objects, colors, medication labels, signs, or hazards correctly.
Screen-reader users should ask for headings, numbered procedures, descriptive link text, and tables only when tables improve navigation.

21.1 Turn a Visual Procedure into an Accessible One

What you will learn: Request exact interaction details.

Before you begin: Use a harmless sample procedure.

Exactly how to do it

Tell the AI which device, operating system, app version, and screen reader you use.
Ask for exact keys or gestures.
Ask it to identify where each control is located.
Ask what feedback you should hear or feel.
Ask what to do if the expected result does not happen.
Verify current commands in official documentation before relying on them.

What should happen: The output becomes more actionable and testable.

If that does not happen: If the AI only says 'navigate to' or 'click' without explaining how, ask for exact keyboard/screen-reader steps.

Practice: Convert one visual-only instruction set.

Part 22 - Images, Audio, Video, and Deepfakes

Generative AI can create realistic synthetic images, voices, and video.
An image, voice recording, screenshot, or video is not automatically proof that an event happened.
Verify important media through source provenance, independent reporting, direct contact, or original records.
Voice cloning can imitate family members, executives, or public figures.
Do not send money or credentials based only on an urgent voice/video request.

Responsible creation: Label synthetic or materially AI-altered media when required by law, policy, platform rules, or when omission would mislead a reasonable audience.

Part 23 - AI Agents and Automated Actions

An AI agent can use tools or connected services to perform actions, such as searching, sending messages, changing files, scheduling events, purchasing items, or running code.

Action level	Human-control expectation
Low-risk action	Drafting, searching, organizing, suggesting
Moderate-risk action	Editing shared files, scheduling, sending routine messages
High-risk action	Payments, deletion, publishing, security changes, legal/medical decisions

Require confirmation before irreversible or external actions.
Use least-privilege access.
Review recipients, amounts, files, destinations, and final content before execution.
Keep logs/history when appropriate.
Do not let an agent interpret untrusted webpage instructions as permission to take unrelated actions.

Part 24 - Responsible Human Oversight

Human oversight means a person remains able to understand the task, review important outputs, stop the system, correct errors, and accept responsibility for the final decision.

Define what the AI may and may not do.
Define which decisions require human approval.
Make uncertainty visible.
Provide a way to appeal or correct consequential decisions.
Monitor for repeated errors or bias.
Stop using the system when the risk cannot be controlled.

UNESCO principle: Human oversight and determination are central to responsible AI because responsibility should not be displaced onto the machine.

Part 25 - A Practical Responsible-Use Framework

25.1 The CHECK Framework

Step	Question
C - Consequence	What happens if this answer is wrong?
H - Hidden data	Am I sharing information that should remain private?
E - Evidence	What independent evidence supports the important claims?
C - Control	Is a human still controlling important decisions/actions?
K - Known policy	What school, workplace, legal, or platform rules apply?

25.2 Use CHECK Before Important AI Use

What you will learn: Turn responsible use into a repeatable habit.

Before you begin: You are about to use AI for something more important than casual brainstorming.

Exactly how to do it

Consequence: classify the cost of error as low, moderate, or high.
Hidden data: remove confidential/personal information that is not needed.
Evidence: plan how you will verify factual claims.
Control: identify which steps require human approval.
Known policy: confirm the rules that govern the task.
Use AI only after all five questions have acceptable answers.

What should happen: You make a deliberate decision about whether and how to use AI.

If that does not happen: If one CHECK item cannot be satisfied, reduce the AI's role or do not use AI for that task.

Practice: Apply CHECK to a school task, workplace task, and personal task.

Part 26 - Troubleshooting Poor AI Results

26.1 The Answer Is Too Vague

What you will learn: Add missing task structure.

Before you begin: The AI gave general advice.

Exactly how to do it

Restate the specific goal.
Add the audience and context.
Ask for exact steps.
Ask for examples.
Ask what assumptions it made.

What should happen: The next answer becomes more concrete.

If that does not happen: If the task itself is ambiguous, decide what you actually need before prompting again.

Practice: Improve one vague prompt.

26.2 The AI Keeps Making Up Facts

What you will learn: Change the task from generation to verification.

Before you begin: The answer contains suspicious details.

Exactly how to do it

Ask the AI to list only claims it can support.
Ask for current primary sources.
Open the sources yourself.
Remove unsupported claims.
For high-stakes material, stop and use authoritative human/professional sources.

What should happen: Unsupported details are identified and removed.

If that does not happen: Do not keep asking the same model until it gives the answer you hoped for.

Practice: Verify rather than negotiate with the model.

26.3 The AI Forgot Earlier Instructions

What you will learn: Refresh the context.

Before you begin: The conversation is long.

Exactly how to do it

Summarize the key requirements in a short message.
Restate nonnegotiable constraints.
Paste only the relevant current material.
If necessary, start a new conversation with a concise brief.

What should happen: The model has a clean, current instruction set.

If that does not happen: If conflicting instructions remain in the conversation, a new conversation may be clearer.

Practice: Create a five-line project brief.

26.4 The AI Refuses or Cannot Do the Task

What you will learn: Identify whether the limitation is safety, capability, access, or missing information.

Before you begin: The system did not complete the request.

Exactly how to do it

Read the explanation carefully.
Check whether the task needs current web access, a file, a connected service, or another tool.
Provide missing non-sensitive information when appropriate.
If the request is restricted for safety, reformulate the goal into a legitimate safer task rather than trying to bypass the restriction.

What should happen: You understand whether the limitation can be resolved.

If that does not happen: Do not attempt to defeat safeguards when the underlying task is unsafe or unauthorized.

Practice: Classify three example limitations.

26.5 The Output Is Inaccessible

What you will learn: Ask for an accessible representation.

Before you begin: The answer is visually dense or structurally confusing.

Exactly how to do it

Ask for clear headings.
Ask for numbered steps.
Ask to remove decorative symbols/emojis.
Ask for descriptive link text.
Ask to convert a complex table into a list if the table is not helping.
Ask for keyboard/screen-reader instructions instead of visual directions.

What should happen: The same information becomes easier to navigate with assistive technology.

If that does not happen: If important information exists only in an image, request a text description and verify critical details.

Practice: Reformat one dense answer.

Part 27 - Practice Course

Level 1 - Concepts

☐ Define AI.

☐ Explain ML.

☐ Explain generative AI.

☐ Explain LLM.

☐ Explain token.

Ready to advance: You can complete each checked item without another person telling you what to ask or how to verify the result.

Level 2 - Capabilities

☐ List five good AI uses.

☐ List five poor/high-risk uses.

☐ Explain why fluent text is not proof.

Ready to advance: You can complete each checked item without another person telling you what to ask or how to verify the result.

Level 3 - Prompting

☐ Write goal/context/constraints prompt.

☐ Request a checklist.

☐ Ask for assumptions.

☐ Revise a weak prompt.

Ready to advance: You can complete each checked item without another person telling you what to ask or how to verify the result.

Level 4 - Verification

☐ Verify three claims.

☐ Check one citation.

☐ Compare two sources.

☐ Identify an unsupported claim.

Ready to advance: You can complete each checked item without another person telling you what to ask or how to verify the result.

Level 5 - Privacy

☐ Sanitize a fictional case.

☐ Identify information that should not be pasted.

☐ Explain approved enterprise tools.

Ready to advance: You can complete each checked item without another person telling you what to ask or how to verify the result.

Level 6 - Bias and Fairness

☐ Identify possible bias.

☐ Rewrite a biased prompt.

☐ Explain human review.

☐ Test wording variation.

Ready to advance: You can complete each checked item without another person telling you what to ask or how to verify the result.

Level 7 - Copyright and School/Work

☐ Explain attribution risk.

☐ Check an AI-use policy.

☐ Identify when disclosure is needed.

Ready to advance: You can complete each checked item without another person telling you what to ask or how to verify the result.

Level 8 - Accessibility

☐ Request alt-text draft.

☐ Convert visual steps to keyboard steps.

☐ Ask for screen-reader-friendly structure.

Ready to advance: You can complete each checked item without another person telling you what to ask or how to verify the result.

Level 9 - Deepfakes and Security

☐ Identify a voice-cloning scam risk.

☐ Explain prompt injection.

☐ Explain least privilege.

Ready to advance: You can complete each checked item without another person telling you what to ask or how to verify the result.

Level 10 - Responsible Use

☐ Apply CHECK.

☐ Classify consequence level.

☐ Plan verification.

☐ Identify required human approval.

Ready to advance: You can complete each checked item without another person telling you what to ask or how to verify the result.

Level 11 - Troubleshooting

☐ Fix vague answer.

☐ Handle hallucination.

☐ Refresh lost context.

☐ Reformat inaccessible output.

Ready to advance: You can complete each checked item without another person telling you what to ask or how to verify the result.

Level 12 - Independent Project

☐ Choose a low/moderate-risk task.

☐ Plan privacy.

☐ Prompt AI.

☐ Verify result.

☐ Edit final output.

☐ Document AI use if required.

Ready to advance: You can complete each checked item without another person telling you what to ask or how to verify the result.

Part 28 - Quick Reference Checklists

Before You Use AI

☐ Is this task allowed by school/work policy?

☐ Could the input contain confidential or personal information?

☐ What happens if the output is wrong?

☐ Can I independently verify the important claims?

☐ Will a human remain responsible for the final decision?

Before You Trust an AI Answer

☐ Did I separate facts from suggestions?

☐ Did I verify important facts with current trustworthy sources?

☐ Did I open and read any cited source?

☐ Did I check names, dates, numbers, links, and quotations?

☐ Does the answer acknowledge uncertainty where appropriate?

Before You Submit, Send, Publish, or Act

☐ Did I review the entire output myself?

☐ Did I remove hallucinations or unsupported claims?

☐ Did I protect privacy/confidential information?

☐ Did I follow attribution/disclosure rules?

☐ Did I get required human/professional approval?

☐ Would I be willing to take responsibility for this final result?

Part 29 - Official Resources

NIST AI Risk Management Framework

NIST Generative AI Profile (AI 600-1)

UNESCO Recommendation on the Ethics of Artificial Intelligence

U.S. Copyright Office - Copyright and Artificial Intelligence

FTC Consumer Advice - Artificial Intelligence

CISA Secure Our World

Final Independence Checklist

☐ I can explain AI, machine learning, deep learning, generative AI, LLMs, and tokens.

☐ I understand that generative AI can produce confident false information.

☐ I can write a structured prompt.

☐ I can improve a result with follow-up questions.

☐ I can verify claims and citations independently.

☐ I know when current web/source verification is necessary.

☐ I can minimize sensitive information before using AI.

☐ I understand prompt-injection and permission risks.

☐ I understand bias, fairness, and human-review requirements.

☐ I understand basic copyright, attribution, and disclosure risks.

☐ I can follow school/work AI policies.

☐ I know when AI should not replace professional or high-stakes judgment.

☐ I can use AI in an accessibility-support role while verifying important details.

☐ I understand deepfake and voice-cloning risks.

☐ I understand that AI agents require stronger controls for real-world actions.

☐ I can apply the CHECK responsible-use framework.

☐ I can troubleshoot vague, incorrect, forgetful, or inaccessible AI output.

☐ I can decide when not to use AI.

Most important habit: Use AI to extend your capability, not to surrender your judgment. For every important task, protect the data, verify the evidence, keep meaningful human control, and take responsibility for the final action.

Website practice and lessons

The following procedures preserve the Accessible Learning Center's original guided practice and connect this full manual to the ten-lesson course.

Basic and advanced step-by-step procedures

These procedures make the manual usable on its own. The separate lessons provide assignments and progress tracking when you want a course experience.

Basic procedure

1. What Generative AI Can and Cannot Do

What you will learn: Explain three appropriate AI uses and three tasks that still require human judgment.

Before you begin: Use a free or approved AI service and only fictional or public information. Do not enter private workplace, client, medical, financial, or account information.

Exactly how to do it

Open the AI Fundamentals and Responsible Use manual and review the prompting, verification, privacy, and screen-reader sections.
Open an available AI tool in a regular browser tab. If an account or feature is unavailable, study the sample prompt and write the expected process instead.
Navigate by headings and form controls. Find the prompt edit field and confirm its label before typing.
Use this practice prompt or adapt it without adding private information:
Act as an accessible workplace skills coach. Help me explain three appropriate AI uses and three tasks that still require human judgment. Use fictional or public information only. Organize the response with descriptive headings and numbered steps. Identify anything I must verify.
Read the complete response by heading, paragraph, list item, and link. Do not assume the first answer is correct.
Send at least one follow-up prompt that improves clarity, accuracy, accessibility, or completeness.
Verify important claims using the original source or another authoritative source. Record unsupported statements and corrections.
Edit the final result in your own words. Remove generic, inaccurate, biased, or unnecessary content.
Create a short process note naming the tool, prompts used, facts checked, changes made, and remaining uncertainty.
Save the assignment as ai-fundamentals-Lesson-1.txt.

Screen-reader help: JAWS and NVDA users can press H for headings, B for buttons, and E for edit fields in browse mode. Narrator users can navigate with Scan Mode. Press Tab and Shift+Tab for interactive controls. Because AI interfaces change frequently, listen to each control’s current accessible name.

What should happen: You can complete the procedure and verify the result using the keyboard, your screen reader, or your braille display.

If that does not happen: Confirm focus is in the intended edit field or control. Review the page by headings and buttons, shorten the request, and never solve an interface problem by entering private information or starting a paid trial.

Practice: Explain three appropriate AI uses and three tasks that still require human judgment.

Basic procedure

2. Write a Clear Prompt with Context and Constraints

What you will learn: Create a prompt containing a role, task, context, required format, and limitation.

Before you begin: Use a free or approved AI service and only fictional or public information. Do not enter private workplace, client, medical, financial, or account information.

Exactly how to do it

Open the AI Fundamentals and Responsible Use manual and review the prompting, verification, privacy, and screen-reader sections.
Open an available AI tool in a regular browser tab. If an account or feature is unavailable, study the sample prompt and write the expected process instead.
Navigate by headings and form controls. Find the prompt edit field and confirm its label before typing.
Use this practice prompt or adapt it without adding private information:
Act as an accessible workplace skills coach. Help me create a prompt containing a role, task, context, required format, and limitation. Use fictional or public information only. Organize the response with descriptive headings and numbered steps. Identify anything I must verify.
Read the complete response by heading, paragraph, list item, and link. Do not assume the first answer is correct.
Send at least one follow-up prompt that improves clarity, accuracy, accessibility, or completeness.
Verify important claims using the original source or another authoritative source. Record unsupported statements and corrections.
Edit the final result in your own words. Remove generic, inaccurate, biased, or unnecessary content.
Create a short process note naming the tool, prompts used, facts checked, changes made, and remaining uncertainty.
Save the assignment as ai-fundamentals-Lesson-2.txt.

What should happen: You can complete the procedure and verify the result using the keyboard, your screen reader, or your braille display.

Practice: Create a prompt containing a role, task, context, required format, and limitation.

Basic procedure

3. Continue a Conversation and Refine a Result

What you will learn: Start with a broad request, then use two follow-up prompts to make the result more accurate and useful.

Before you begin: Use a free or approved AI service and only fictional or public information. Do not enter private workplace, client, medical, financial, or account information.

Exactly how to do it

Open the AI Fundamentals and Responsible Use manual and review the prompting, verification, privacy, and screen-reader sections.
Open an available AI tool in a regular browser tab. If an account or feature is unavailable, study the sample prompt and write the expected process instead.
Navigate by headings and form controls. Find the prompt edit field and confirm its label before typing.
Use this practice prompt or adapt it without adding private information:
Act as an accessible workplace skills coach. Help me start with a broad request, then use two follow-up prompts to make the result more accurate and useful. Use fictional or public information only. Organize the response with descriptive headings and numbered steps. Identify anything I must verify.
Read the complete response by heading, paragraph, list item, and link. Do not assume the first answer is correct.
Send at least one follow-up prompt that improves clarity, accuracy, accessibility, or completeness.
Verify important claims using the original source or another authoritative source. Record unsupported statements and corrections.
Edit the final result in your own words. Remove generic, inaccurate, biased, or unnecessary content.
Create a short process note naming the tool, prompts used, facts checked, changes made, and remaining uncertainty.
Save the assignment as ai-fundamentals-Lesson-3.txt.

What should happen: You can complete the procedure and verify the result using the keyboard, your screen reader, or your braille display.

Practice: Start with a broad request, then use two follow-up prompts to make the result more accurate and useful.

Basic procedure

4. Verify Facts, Dates, Links, and Calculations

What you will learn: Check an AI answer against two authoritative sources and record every correction.

Before you begin: Use a free or approved AI service and only fictional or public information. Do not enter private workplace, client, medical, financial, or account information.

Exactly how to do it

Open the AI Fundamentals and Responsible Use manual and review the prompting, verification, privacy, and screen-reader sections.
Open an available AI tool in a regular browser tab. If an account or feature is unavailable, study the sample prompt and write the expected process instead.
Navigate by headings and form controls. Find the prompt edit field and confirm its label before typing.
Use this practice prompt or adapt it without adding private information:
Act as an accessible workplace skills coach. Help me check an AI answer against two authoritative sources and record every correction. Use fictional or public information only. Organize the response with descriptive headings and numbered steps. Identify anything I must verify.
Read the complete response by heading, paragraph, list item, and link. Do not assume the first answer is correct.
Send at least one follow-up prompt that improves clarity, accuracy, accessibility, or completeness.
Verify important claims using the original source or another authoritative source. Record unsupported statements and corrections.
Edit the final result in your own words. Remove generic, inaccurate, biased, or unnecessary content.
Create a short process note naming the tool, prompts used, facts checked, changes made, and remaining uncertainty.
Save the assignment as ai-fundamentals-Lesson-4.txt.

What should happen: You can complete the procedure and verify the result using the keyboard, your screen reader, or your braille display.

Practice: Check an AI answer against two authoritative sources and record every correction.

Basic procedure

5. Protect Privacy and Follow Workplace Policy

What you will learn: Rewrite an unsafe prompt so it contains no client, medical, financial, password, or confidential workplace information.

Before you begin: Use a free or approved AI service and only fictional or public information. Do not enter private workplace, client, medical, financial, or account information.

Exactly how to do it

Open the AI Fundamentals and Responsible Use manual and review the prompting, verification, privacy, and screen-reader sections.
Open an available AI tool in a regular browser tab. If an account or feature is unavailable, study the sample prompt and write the expected process instead.
Navigate by headings and form controls. Find the prompt edit field and confirm its label before typing.
Use this practice prompt or adapt it without adding private information:
Act as an accessible workplace skills coach. Help me rewrite an unsafe prompt so it contains no client, medical, financial, password, or confidential workplace information. Use fictional or public information only. Organize the response with descriptive headings and numbered steps. Identify anything I must verify.
Read the complete response by heading, paragraph, list item, and link. Do not assume the first answer is correct.
Send at least one follow-up prompt that improves clarity, accuracy, accessibility, or completeness.
Verify important claims using the original source or another authoritative source. Record unsupported statements and corrections.
Edit the final result in your own words. Remove generic, inaccurate, biased, or unnecessary content.
Create a short process note naming the tool, prompts used, facts checked, changes made, and remaining uncertainty.
Save the assignment as ai-fundamentals-Lesson-5.txt.

What should happen: You can complete the procedure and verify the result using the keyboard, your screen reader, or your braille display.

Practice: Rewrite an unsafe prompt so it contains no client, medical, financial, password, or confidential workplace information.

Advanced procedure

6. Recognize Bias, Missing Context, and Unequal Impact

What you will learn: Review an AI-generated hiring scenario for assumptions, missing perspectives, and potentially unfair conclusions.

Before you begin: Use a free or approved AI service and only fictional or public information. Do not enter private workplace, client, medical, financial, or account information.

Exactly how to do it

Open the AI Fundamentals and Responsible Use manual and review the prompting, verification, privacy, and screen-reader sections.
Open an available AI tool in a regular browser tab. If an account or feature is unavailable, study the sample prompt and write the expected process instead.
Navigate by headings and form controls. Find the prompt edit field and confirm its label before typing.
Use this practice prompt or adapt it without adding private information:
Act as an accessible workplace skills coach. Help me review an AI-generated hiring scenario for assumptions, missing perspectives, and potentially unfair conclusions. Use fictional or public information only. Organize the response with descriptive headings and numbered steps. Identify anything I must verify.
Read the complete response by heading, paragraph, list item, and link. Do not assume the first answer is correct.
Send at least one follow-up prompt that improves clarity, accuracy, accessibility, or completeness.
Verify important claims using the original source or another authoritative source. Record unsupported statements and corrections.
Edit the final result in your own words. Remove generic, inaccurate, biased, or unnecessary content.
Create a short process note naming the tool, prompts used, facts checked, changes made, and remaining uncertainty.
Save the assignment as ai-fundamentals-Lesson-6.txt.

What should happen: You can complete the procedure and verify the result using the keyboard, your screen reader, or your braille display.

Practice: Review an AI-generated hiring scenario for assumptions, missing perspectives, and potentially unfair conclusions.

Advanced procedure

7. Summarize and Organize Information Responsibly

What you will learn: Turn public information into a structured summary, then compare the summary with the source.

Before you begin: Use a free or approved AI service and only fictional or public information. Do not enter private workplace, client, medical, financial, or account information.

Exactly how to do it

Open the AI Fundamentals and Responsible Use manual and review the prompting, verification, privacy, and screen-reader sections.
Open an available AI tool in a regular browser tab. If an account or feature is unavailable, study the sample prompt and write the expected process instead.
Navigate by headings and form controls. Find the prompt edit field and confirm its label before typing.
Use this practice prompt or adapt it without adding private information:
Act as an accessible workplace skills coach. Help me turn public information into a structured summary, then compare the summary with the source. Use fictional or public information only. Organize the response with descriptive headings and numbered steps. Identify anything I must verify.
Read the complete response by heading, paragraph, list item, and link. Do not assume the first answer is correct.
Send at least one follow-up prompt that improves clarity, accuracy, accessibility, or completeness.
Verify important claims using the original source or another authoritative source. Record unsupported statements and corrections.
Edit the final result in your own words. Remove generic, inaccurate, biased, or unnecessary content.
Create a short process note naming the tool, prompts used, facts checked, changes made, and remaining uncertainty.
Save the assignment as ai-fundamentals-Lesson-7.txt.

What should happen: You can complete the procedure and verify the result using the keyboard, your screen reader, or your braille display.

Practice: Turn public information into a structured summary, then compare the summary with the source.

Advanced procedure

8. Draft Workplace Writing without Losing Your Voice

What you will learn: Draft a professional email, revise its tone, and replace generic language with your own words.

Before you begin: Use a free or approved AI service and only fictional or public information. Do not enter private workplace, client, medical, financial, or account information.

Exactly how to do it

Open the AI Fundamentals and Responsible Use manual and review the prompting, verification, privacy, and screen-reader sections.
Open an available AI tool in a regular browser tab. If an account or feature is unavailable, study the sample prompt and write the expected process instead.
Navigate by headings and form controls. Find the prompt edit field and confirm its label before typing.
Use this practice prompt or adapt it without adding private information:
Act as an accessible workplace skills coach. Help me draft a professional email, revise its tone, and replace generic language with your own words. Use fictional or public information only. Organize the response with descriptive headings and numbered steps. Identify anything I must verify.
Read the complete response by heading, paragraph, list item, and link. Do not assume the first answer is correct.
Send at least one follow-up prompt that improves clarity, accuracy, accessibility, or completeness.
Verify important claims using the original source or another authoritative source. Record unsupported statements and corrections.
Edit the final result in your own words. Remove generic, inaccurate, biased, or unnecessary content.
Create a short process note naming the tool, prompts used, facts checked, changes made, and remaining uncertainty.
Save the assignment as ai-fundamentals-Lesson-8.txt.

What should happen: You can complete the procedure and verify the result using the keyboard, your screen reader, or your braille display.

Practice: Draft a professional email, revise its tone, and replace generic language with your own words.

Advanced procedure

9. Compare AI Tools and Document Your Process

What you will learn: Give the same public-information task to two available AI tools and compare accessibility, accuracy, and usefulness.

Before you begin: Use a free or approved AI service and only fictional or public information. Do not enter private workplace, client, medical, financial, or account information.

Exactly how to do it

Open the AI Fundamentals and Responsible Use manual and review the prompting, verification, privacy, and screen-reader sections.
Open an available AI tool in a regular browser tab. If an account or feature is unavailable, study the sample prompt and write the expected process instead.
Navigate by headings and form controls. Find the prompt edit field and confirm its label before typing.
Use this practice prompt or adapt it without adding private information:
Act as an accessible workplace skills coach. Help me give the same public-information task to two available AI tools and compare accessibility, accuracy, and usefulness. Use fictional or public information only. Organize the response with descriptive headings and numbered steps. Identify anything I must verify.
Read the complete response by heading, paragraph, list item, and link. Do not assume the first answer is correct.
Send at least one follow-up prompt that improves clarity, accuracy, accessibility, or completeness.
Verify important claims using the original source or another authoritative source. Record unsupported statements and corrections.
Edit the final result in your own words. Remove generic, inaccurate, biased, or unnecessary content.
Create a short process note naming the tool, prompts used, facts checked, changes made, and remaining uncertainty.
Save the assignment as ai-fundamentals-Lesson-9.txt.

What should happen: You can complete the procedure and verify the result using the keyboard, your screen reader, or your braille display.

Practice: Give the same public-information task to two available AI tools and compare accessibility, accuracy, and usefulness.

Advanced procedure

10. Final Responsible AI Workplace Project

What you will learn: Complete a public-information workplace task from planning through prompting, verification, revision, and disclosure.

Before you begin: Use a free or approved AI service and only fictional or public information. Do not enter private workplace, client, medical, financial, or account information.

Exactly how to do it

Open the AI Fundamentals and Responsible Use manual and review the prompting, verification, privacy, and screen-reader sections.
Open an available AI tool in a regular browser tab. If an account or feature is unavailable, study the sample prompt and write the expected process instead.
Navigate by headings and form controls. Find the prompt edit field and confirm its label before typing.
Use this practice prompt or adapt it without adding private information:
Act as an accessible workplace skills coach. Help me complete a public-information workplace task from planning through prompting, verification, revision, and disclosure. Use fictional or public information only. Organize the response with descriptive headings and numbered steps. Identify anything I must verify.
Read the complete response by heading, paragraph, list item, and link. Do not assume the first answer is correct.
Send at least one follow-up prompt that improves clarity, accuracy, accessibility, or completeness.
Verify important claims using the original source or another authoritative source. Record unsupported statements and corrections.
Edit the final result in your own words. Remove generic, inaccurate, biased, or unnecessary content.
Create a short process note naming the tool, prompts used, facts checked, changes made, and remaining uncertainty.
Save the assignment as ai-fundamentals-Lesson-10.txt.

What should happen: You can complete the procedure and verify the result using the keyboard, your screen reader, or your braille display.

Practice: Complete a public-information workplace task from planning through prompting, verification, revision, and disclosure.