AI Without the Hype
Sources, Standards, and Claim Boundaries
The evidence ledger behind UVLM’s beginner AI course: official definitions, risk-management guidance, education frameworks, verification methods, privacy resources, scam warnings, labor research, copyright guidance, and UVLM’s separately labeled educational framework.
A citation shows where a statement came from. It does not make every interpretation correct. Sources must still be read in context, checked for date and scope, and revised when the underlying guidance changes.
How to read this page
Not every source has the same authority or purpose.
The labels below prevent official guidance, research methods, UVLM proposals, developing drafts, and unresolved questions from being blended into one undifferentiated voice.
Governmental or intergovernmental source
Used for definitions, risk guidance, consumer warnings, public policy, or current institutional recommendations within that source’s scope.
Research-informed teaching framework
Used for lesson design, verification practice, accessibility, learner agency, or facilitation—not as proof of every factual claim about AI.
Internal or UVLM-authored framework
Used as design lineage or an educational lens. It is not presented as independent validation or scientific consensus.
Developing guidance
Useful for monitoring direction, but not treated as a final standard. The page records the status reviewed on the date shown.
Contested or unresolved issue
The course may explain the debate but should not manufacture certainty about consciousness, future labor effects, or other unsettled matters.
Responsible institution
An agency, employer, school, court, provider, or organization is generally the first source for its current procedure, settings, or official requirement.
Source-selection method
Match the evidence to the claim.
Prefer responsible primary sources
Use the institution that owns a current rule, procedure, warning, standard, or public record before relying on a generated summary.
Use independent sources for evaluation
A company’s documentation may explain its own product, but independent testing or research is needed for broader performance and impact claims.
Check date, scope, and status
A real source can still be outdated, quoted too broadly, limited to another population, or superseded by a revision.
Preserve unresolved findings
“Unresolved,” “partially supported,” and “professional review required” are valid outcomes. The source process should not force every question into yes or no.
External source register
Official and research-informed foundations.
Select a category to view the sources, what they support, and the limits on how this course uses them.
01 AI Definitions, Human-AI Tasks, and Risk Management
OECD — Updated Definition of an AI System
Provides an internationally recognized, technology-neutral definition and explanatory memorandum. It supports the course’s treatment of AI as a broad family of systems rather than one chatbot or one company.
Course use: introductory definition, system-level scope, and differentiation of outputs such as predictions, content, recommendations, and decisions.
Open the OECD explanatory memorandum →NIST AI Risk Management Framework 1.0
A voluntary, rights-preserving, use-case-agnostic framework for managing AI risks to individuals, organizations, and society.
Course use: lifecycle thinking, governance, context mapping, measurement, management, and the principle that trustworthiness must be evaluated rather than assumed.
Status note: NIST states that AI RMF 1.0 is being revised.
Open the NIST AI RMF hub →NIST AI 600-1 — Generative AI Profile
Companion guidance identifying generative-AI risks and suggested risk-management actions. It discusses risks such as confabulation, information integrity, harmful bias, human-AI configuration, privacy, security, and content provenance.
Course use: why fluent output can be false, why overreliance matters, and why evaluation must include the human and organizational context.
Open NIST AI 600-1 →NIST — AI Use Taxonomy: A Human-Centered Approach
Classifies human-AI activities by the role a system performs in achieving a human outcome, independent of one technique or domain.
Course use: ask what job the system performs, what human goal it serves, and what must be measured for that particular activity.
Open the NIST AI Use Taxonomy →02 Public AI Literacy, Education, Libraries, and Accessibility
American Library Association — Guidance on AI in Libraries
ALA’s 2026 guidance centers public good, intellectual freedom, privacy, sustainability, access, labor, meaningful human agency, community input, and a harm-reduction approach to AI literacy.
Course use: meet patrons where they are, avoid shame, teach verification and privacy, preserve human assistance, and give reviewers authority to correct, reject, or escalate AI output.
Open ALA’s adopted guidance →UNESCO — Generative AI in Education and Research
UNESCO’s guidance promotes a human-centered approach, inclusion, equity, privacy protection, critical thinking, creativity, and institutional validation of pedagogical and ethical suitability.
Course use: AI should support human agency and learning rather than displace teachers, relationships, or critical thought.
Open UNESCO’s AI and Futures of Learning hub →UNESCO — AI Competency Frameworks for Students and Teachers
The 2024 frameworks emphasize human-centered mindsets, ethics, foundational understanding, responsible application, pedagogy, design, and lifelong development.
Course use: understand what AI can and cannot do, retain agency, apply ethics, and build competencies across levels rather than teaching one product’s buttons.
Open UNESCO’s framework links →CAST — Universal Design for Learning Guidelines 3.0
UDL 3.0 supports learning environments that reduce barriers and build learner agency through multiple means of engagement, representation, and action or expression.
Course use: observation choices, multiple formats, one-step scaffolding, reflection, accessible materials, and no assumption that one device or learning mode fits everyone.
Open CAST UDL Guidelines 3.0 →Digital Inquiry Group — Civic Online Reasoning
Provides evidence-based instruction for evaluating online information through questions such as who is behind the information, what the evidence is, and what other sources say. It also teaches lateral reading and click restraint.
Course use: leave an unfamiliar source, investigate it elsewhere, open original evidence, and compare independent coverage.
Open the Civic Online Reasoning posters →03 Privacy, Data Minimization, and De-Identification
NIST Privacy Framework
A voluntary framework for identifying and managing privacy risk. Version 1.0 remains available; NIST’s Version 1.1 page was still marked “coming soon” when this source register was reviewed.
Course use: privacy is a risk-management lifecycle, not only a password issue; organizations must examine collection, use, retention, sharing, control, and consequences.
Open the NIST Privacy Framework hub →NIST Digital Identity Guidelines — Data Minimization
NIST’s privacy considerations state that systems should collect and process only personal information necessary for the defined identity purpose and warn that unnecessary processing or retention may cause loss of autonomy, trust, or security.
Course use: ask for the minimum information required and begin with a general description or placeholder.
Open NIST privacy considerations →NIST — De-Identification Guidance
NIST explains that de-identification can reduce privacy risk but that some de-identified information can be re-identified. Later guidance also emphasizes governance and limitations of traditional approaches.
Course use: removing a name helps but does not guarantee anonymity; combinations of details may still identify a person.
Open NIST SP 800-188 information →ALA — Patron Privacy and AI Vendor Review
ALA recommends examining what AI systems collect, log, retain, share, delete, and use for model improvement, while prohibiting entry of personally identifiable and nonpublic patron records unless a system has been formally approved.
Course use: no private classroom demonstrations, no patron records in consumer tools, and a meaningful non-AI path when possible.
Open ALA privacy guidance →04 Confabulation, Verification, and AI-Content Detection
NIST Generative AI Profile — Confabulation and Information Integrity
Supports the course’s distinction between plausible generation and verified knowledge, and its treatment of overreliance, human-AI configuration, and information integrity as risk-management concerns.
Course use: facts, assumptions, uncertainty, and recommendations should remain visible and separately reviewable.
Open NIST AI 600-1 →NIST GenAI Evaluations
NIST evaluates generators, detectors, and prompting across text, image, code, audio, and video. Its published program notes a performance gap between generation and detection and reports that strong generated summaries fooled every detector in an early pilot.
Course use: an AI detector is one clue, not conclusive proof of authorship or authenticity.
Open NIST GenAI evaluations →Civic Online Reasoning — Lateral Reading
Supports leaving an unfamiliar website to see what trusted sources say about it rather than relying on the site’s own description of its authority.
Course use: verification should change the source, reviewer, method, or evidence—not ask the same answer to certify itself.
Open the lateral-reading resource →05 Scams, Voice Cloning, Deepfakes, and Impersonation
FTC — AI-Enhanced Family Emergency Scams
Warns that scammers may clone a loved one’s voice from a short audio clip. FTC guidance says not to trust the voice alone and to call the person through a number already known to be theirs.
Course use: independent verification, refusal of unusual payment methods, and a no-shame family safety agreement.
Open FTC voice-cloning guidance →FBI IC3 — AI-Generated Videos and Spoofed Government Sites
The July 20, 2026 alert describes scammers impersonating IC3, using AI-generated videos, spoofed websites, fear, urgency, and false recovery services to revictimize people.
Course use: appearance is not proof; verify through known official channels and do not pay supposed recovery agents.
Open the July 2026 FBI IC3 alert →FTC — Government Impersonator and Recovery Scams
FTC’s June 2026 warning explains that real FTC employees do not contact people through text or messaging apps to recover scam losses, demand payment, or request financial information.
Course use: understand what a real institution will not do and independently locate contact information.
Open the FTC impersonator warning →Official Fraud-Reporting Resources
FTC ReportFraud supports U.S. consumer fraud reports; IdentityTheft.gov provides identity-theft recovery planning; the FBI IC3 accepts cyber-enabled crime complaints.
06 Work, Occupational Exposure, and Human Expertise
ILO–NASK — Generative AI and Jobs, 2025
The refined global index analyzes occupational exposure at the task level and concludes that one in four jobs is potentially exposed to generative AI, while transformation rather than complete replacement is the more likely overall outcome.
Course use: avoid both panic and false reassurance; examine actual tasks, job quality, worker autonomy, training, industry, and local conditions.
Open the ILO–NASK report page →ALA — Labor, Expertise, and Uncompensated Review
ALA’s guidance calls for worker consultation, preservation of professional expertise, meaningful human review, and use of verified efficiency gains to improve work and services rather than justify deskilling or uncompensated review burdens.
Course use: count shifted workload and ask who inherits the review burden when AI creates more output.
Open ALA labor guidance →07 Creative Work, Digital Replicas, Copyrightability, and Training
Copyright and Artificial Intelligence Report
The U.S. Copyright Office’s multi-part study covers digital replicas, copyrightability of generative-AI outputs, and generative-AI training.
Course use: legal questions vary by the human contribution, source material, use, license, and jurisdiction; generated output should not be treated as automatically owned or automatically free of rights issues.
Open the Copyright Office AI study →Copyrightability of Generative-AI Outputs
The Copyright Office states that AI assistance does not bar copyright protection, but protection requires sufficient human-authored expressive elements. Mere prompting alone is generally insufficient.
Course use: preserve records of human authorship, editing, arrangement, source permission, and disclosure; seek legal advice for important publication or commercial decisions.
Open the Part 2 announcement →UVLM source lineage
Educational architecture—labeled as ours.
The sources below explain where UVLM-specific concepts came from. They are internal primary design sources, not independent proof that the concepts are scientifically established.
The Coherence Lattice: A Probabilistic Framework for Unified Inference
Supplies the project’s guiding commitments: epistemic humility, disciplinary equity, ethical symmetry, and guardrails against metaphysical overreach. It defines the UVLM coherence lens around Empathy, Transparency, and related diagnostic variables.
Course contribution: helpful + honest interaction, no flattening of local expertise, and explicit separation of established science from UVLM synthesis.
Claim boundary: translation lattice and research proposal—not a final Theory of Everything or replacement for domain science.
Universal Control Codex (UCC) Supplement
Encodes tasks, authorities, reasoning steps, evidence requirements, validation rules, reporting structure, and escalation policy as explicit, testable artifacts.
Course contribution: CARES prompting and the practice of making task, evidence, validation, output, and human handoff visible.
Claim boundary: research and experimentation scaffold—not automatic legal, regulatory, or professional compliance.
Appendix Q — Quantum Mechanics and GUFT
States that standard quantum mechanics and quantum field theory remain non-negotiable, while GUFT mappings are descriptive, structural analogies rather than alternative dynamics.
Course contribution: do not use quantum language to imply telepathy, consciousness collapse, mystical causation, or moral properties in physics.
Claim boundary: quantum structures are pattern donors—not a license for “quantum” mystification.
Provenance Memory Reservoirs for Governed AI Cognition
Distinguishes memory from mere retention and proposes that remembered artifacts remain traceable, replayable, correctable, revocable, consent-bounded, and resource-aware.
Course contribution: “remembered” does not mean verified, authorized, current, consented, canonized, or true.
Claim boundary: formal design proposal and empirical research program—not a completed hallucination cure or truth-certification system.
Preventing Hyperreal Design Drift
Warns that elegant names, registries, schemas, and governance artifacts can describe a product without delivering useful work.
Course contribution: every educational page, prompt, and resource should perform a real learner function rather than merely signal sophistication.
Claim boundary: a candidate page or scaffold is not a demonstrated outcome; user-visible utility and testing remain necessary.
Multi-Axial Coherence Analysis for Exogenic Off-Loading
Examines whether transferring work to an external system creates a genuine global efficiency improvement or merely shifts entropy, risk, opacity, cost, or harm elsewhere.
Course contribution: AI efficiency should count privacy, verification, correction, environmental burden, uncompensated review, and the people who inherit failure risk.
Claim boundary: interdisciplinary proposed framework; empirical claims require domain-specific evidence and validation.
Course claim ledger
What supports the course’s central teaching claims?
AI is a broad family of systems.
Basis: OECD definition, NIST taxonomy, ALA guidance.
Limit: the exact technology must be identified for the particular product and use.
Generative AI can produce confident false content.
Basis: NIST Generative AI Profile and evaluation work.
Limit: not every answer is false; risk depends on task, source access, model, and review.
Personal information should be minimized.
Basis: NIST privacy guidance and ALA library privacy guidance.
Limit: de-identification reduces risk but may not eliminate re-identification.
Important claims should be checked laterally.
Basis: Civic Online Reasoning and ALA AI-literacy guidance.
Limit: source quality still depends on relevance, date, evidence, independence, and scope.
A familiar voice or face is not sufficient identity proof.
Basis: FTC voice-cloning guidance and FBI IC3 alerts.
Limit: visible artifacts may help, but independent contact remains the stronger control.
Job effects are likely to be uneven and task-based.
Basis: ILO–NASK 2025 occupational-exposure analysis.
Limit: global exposure estimates do not predict one person’s job outcome.
Meaningful human review requires real authority.
Basis: ALA guidance and NIST risk-management principles.
Limit: review quality depends on expertise, evidence, time, and the ability to change the result.
A useful interaction should be helpful and honest.
Basis: UVLM Empathy × Transparency translation.
Limit: not a universal scientific measure of truth, consciousness, morality, or human worth.
CARES can make a task easier to inspect.
Basis: UCC-style explicit task, evidence, validation, and escalation design.
Limit: a better prompt cannot guarantee truth, privacy, or professional competence.
An AI answer is a candidate—not a verdict.
Basis: UVLM non-authority, provenance, and human-agency doctrine.
Limit: human review is also fallible and should be matched to the consequence of error.
Course claim ceiling
What this program does not claim.
No truth certification
A prompt, citation, memory, hash, audit trail, or review record does not automatically make a claim true.
No consciousness verdict
Conversational fluency is not treated as proof of consciousness, feelings, personhood, or moral authority.
No professional substitution
The course does not replace medical care, legal advice, tax advice, financial advice, emergency services, or official adjudication.
No guaranteed privacy
Prompt instructions and settings can reduce risk but cannot create an absolute confidentiality guarantee across all products.
No universal employment forecast
Exposure studies describe patterns and probabilities, not the certain future of one worker, occupation, or community.
No established GUFT consensus
GUFT and CoherenceLattice concepts are labeled UVLM proposals and educational lenses, not accepted replacements for domain science.
No product endorsement
Listing or demonstrating a tool does not certify its accuracy, safety, privacy, accessibility, sustainability, or suitability.
No perfect detector
AI-content detection is not presented as a conclusive authorship or authenticity test.
No final curriculum
The program remains a developing pilot. Community feedback, accessibility review, evidence updates, and observed failures may change the materials.
Corrections and maintenance
A source register should be correctable.
When reporting an issue, identify the page section, statement, source, relevant date, and proposed correction. Do not send private records or urgent requests.
Report a Source Correction
Use this for a broken link, superseded standard, inaccurate summary, or source that no longer supports the stated claim.
Email a Source CorrectionSuggest an Authoritative Source
Include the source title, issuing organization, publication date, link, and the course claim it would support or challenge.
Suggest a SourceChallenge a UVLM Interpretation
Explain where the course goes beyond its evidence, blends source categories, or presents an internal lens too strongly.
Challenge a Claim BoundaryPage record
- Page version: 0.1
- Reviewed through: August 26, 2026
- Program status: developing public-education pilot
- Maintenance rule: current official sources supersede this summary when they conflict
- Correction address: info@ultraverbaluxmentis.org