AI Foundations · Module 04 April 2026
Agents.·Media.·Work.·Rights.

AI in society. The parts that affect you as an everyday user — not as an engineer. What has changed in the last eighteen months, how to recognize it, and what to do.

What we'll cover 02 / 19
Agenda · four shifts

Four questions everyone using AI should have an answer to.

01 · AGENTS

Software that acts

From chatbot to copilot to agent — what changed and what it means for you.

02 · MEDIA

Synthetic reality

Deepfakes, voice clones, and how to stay safe when anything can be faked.

03 · WORK

Your job, changed

What gets automated, what gets amplified, what stays human.

04 · RIGHTS

Rules catching up

EU AI Act, GDPR, copyright — and the rights they give you.

Part One 03 / 19
01
Part One · The agentic shift

Software that acts.

For a decade, software waited for your click. Then it started drafting. Now it acts. This is the biggest mental-model change of 2025.

Three generations in ten years 04 / 19
2015 → today

Every decade of software, compressed into three years.

2015 · Software you click

Tools

Excel, Salesforce, SAP. You learn the menus. Software waits for instructions.

2023 · Software that drafts

Copilots

Embedded assistants in email, docs, CRM. They suggest. You approve every output.

This module is about the step from the middle column to the right.

The three levels 05 / 19
At a glance

Three levels of autonomy — know which one you're using.

LEVEL 01

Chatbot

Like a receptionist.
You ask, it answers. Nothing happens outside the conversation.
Examples
ChatGPT · Gemini · Claude · website support chat.
LEVEL 02

Copilot

Like an intern next to you.
Drafts the email, the slide, the formula. Every output passes through your review.
Examples
Microsoft 365 Copilot · Gmail "Help me write" · GitHub Copilot.
LEVEL 03

Agent

Like a junior employee with a task.
You set a goal. It plans, uses tools, acts, and returns the finished work.
Examples
ChatGPT agent mode · Claude Computer Use · Salesforce Agentforce.
The one-line shift 06 / 19
Keep this sentence
Chatbotsanswer.
Copilotssuggest.
Agentsdo.
79%
of executives say their organization is adopting AI agents.
PwC · 2025
33%
of enterprise apps will include agents by 2028 — up from under 1% in 2024.
Gartner
>40%
of agent projects will be cancelled by 2027 — governance, not technology, is the hard part.
Gartner
What agents mean for you 07 / 19
Three practical consequences

Agents act in your accounts. You still own the outcome.

01

Pricing shifts

From "per seat" (€20–30 / user / month) to per outcome — per resolved ticket, per processed invoice, per minute of work.

02

Permissions inherit

An agent running on your behalf can reach everything you can reach — including every over-shared folder and "anyone with the link" document.

03

Accountability stays

If the agent misfiles, misrepresents, or misorders — you own it. The human is in the loop whether present or not.

Before you let an agent loose: narrow scope, draft-only mode, audit permissions, define a stop button.

Part Two 08 / 19
02
Part Two · Synthetic media

Anything can now be faked.

A phone, three seconds of audio, and a free app. That's all it takes. The technology is neutral — the implications aren't.

Four flavors of synthetic 09 / 19
What you're up against

Know the shape of the threat.

VIDEO

Deepfake

Face or body swap, lip-sync alignment, full-body motion. Consumer-grade hardware.

AUDIO

Voice clone

Trained on as little as 3 seconds of clean audio. Usable for calls, voicemails, meetings.

IMAGE

Generated photo

Photorealistic, no camera required. DALL·E, Midjourney, Stable Diffusion, Flux.

TEXT

AI at scale

Fake reviews, fake news sites, AI-written spam at industrial volume.

Real incidents · 2024 – 2025 10 / 19
The ledger so far Five flashpoints in eighteen months
Jan 2024
Taylor Swift deepfakes spread to 45M+ views on X before takedown. Platform liability laws revisited worldwide.
Jan 2024
AI-cloned Biden robocall tells New Hampshire voters not to vote. FCC rules AI voice calls illegal under existing law.
Feb 2024
$25M Hong Kong fraud — a finance worker transfers the sum after a deepfaked video call with "the CFO" and colleagues. Multi-person deepfake.
2024
Elections worldwide — coordinated deepfake campaigns in India, EU, US, and South Korea. Democratic trust erodes faster than regulation catches up.
2024 – 25
"Grandchild in trouble" scams rise sharply as voice-cloning tools commoditize. FTC and European banks issue public warnings.
Detection is not the answer 11 / 19
Two uncomfortable facts
Detectors lag generators. And authentic footage becomes deniable.
The liar's dividend · the second-order problem of synthetic media

Humans identify deepfakes correctly around 60% of the time — barely above chance, and worse on high-quality samples. Every detector trains on yesterday's fakes; tomorrow's model is already out. And the real danger isn't only fakes being believed — it's real footage being dismissed because "anyone can fake a video now."

Your everyday playbook 12 / 19
Five rules that actually work

Authenticity must now be proven — not assumed.

01

Second channel

Urgent voice or video asking for money, credentials, approval? Hang up. Call back on a known number.

02

Safe words

Agree on a phrase with family and finance. The real person knows it. A clone won't.

03

Treat unsourced as fiction

Especially when the content is perfectly aligned with what you already want to believe.

04

Check provenance

Content Credentials (C2PA), SynthID, platform labels. Absence isn't proof — presence is signal.

05

Don't seed the clone

Three seconds of clean audio is enough. Public voice posts are training data for someone.

Part Three 13 / 19
03
Part Three · Your work, changed

The real pattern isn't replacement.

Almost no jobs are entirely automatable. Most jobs are a bundle of 10–30 tasks — and AI hits some of them hard, leaves others alone, and amplifies a third group.

The three buckets 14 / 19
Automate · amplify · stays human

Tasks get automated. Jobs change shape.

BUCKET 01

Automated

High-volume, digital, rule-heavy.
AI does the task end-to-end. Human checks or doesn't.
Examples
Tier-1 support · invoice processing · meeting notes · first-draft content · basic translation.
BUCKET 02

Amplified

Expert judgment + AI force multiplier.
These roles get more valuable, not less. AI drafts; human directs and decides.
Examples
Clinicians · lawyers · engineers · designers · analysts · strategists.
BUCKET 03

Stays human

Embodied, accountable, relational, novel.
Where physical presence, high-stakes judgment, or trust cannot be automated away.
Examples
Nursing · trades · leadership · negotiation · teaching kids · therapy · eldercare.
The numbers and the response 15 / 19
What the data says · what you do

The shift from producer to director.

30%
of current US work hours could be automated by 2030.
McKinsey · 2023
+14%
productivity gain for support agents using AI — largest gains for the least experienced workers.
Brynjolfsson · Li · Raymond · 2023
Coding, writing, analysis
consistently top the list of actual AI usage across the economy.
Anthropic Economic Index · 2025

Your response: learn to direct AI (prompting, tool selection, verification). Invest in what amplifies — your judgment, taste, relationships. Audit which of your tasks are becoming "draftable" and redirect time to higher-value work. From "What do I do?" to "What do I decide and review?"

Part Four 16 / 19
04
Part Four · Rights and ownership

Rules catching up.

Slowly, unevenly, and with real teeth in some places. Here's what you need to know as a user — not a compliance officer.

The EU AI Act · what it means for you 17 / 19
Phased in through 2026

The most comprehensive AI regulation in the world — and its reach is global.

Phase Starts What it covers
Prohibited usesFeb 2025Social scoring, manipulative AI, untargeted facial scraping, real-time biometric ID — banned.
General-purpose AIAug 2025Foundation-model providers — transparency, safety testing, copyright disclosures.
High-risk systemsAug 2026Hiring, credit, education, law enforcement — audits, conformance, human oversight.
Full applicationAug 2027Remaining categories, embedded AI in regulated products.

Your practical rights: to know you're talking to an AI · to know content is AI-generated · to human review of high-stakes decisions · to an explanation of decisions affecting you · to lodge a complaint with national AI authorities.

Who owns AI output? 18 / 19
Copyright in the age of GenAI

Nobody — or the human who substantially contributed.

What the rules say
  • US Copyright Office (2023–24) — purely AI-generated works are not copyrightable. Human authorship required.
  • EU — similar standard. Original human contribution required.
  • UK — reduced-term protection exists, but contested for modern GenAI.
  • China — some AI-generated images granted copyright where prompt work is deemed creative.
What it means in practice
  • One-prompt image, no edits — probably not yours to copyright. Anyone can reuse.
  • ChatGPT draft, substantially rewritten by you — your edited version is yours.
  • Commercial use — separate from copyright. Check each tool's Terms of Service.
  • Training-data lawsuits (NYT v. OpenAI, Getty v. Stability) — still being decided. Indemnification offered by Microsoft, Google, Anthropic, Adobe.
End of Module 04 19 / 19
// FIVE EVERYDAY RULES
Verifybefore you forward. Unexpected urgency → second channel.
Disclosewhen AI meaningfully shaped your work.
Minimizewhat you share — only the data the task actually needs.
Match tiersdata sensitivity to tool approval.
Stay in loopfor any consequential decision.

AI is not something happening to you. It's something you participate in — as a worker, a user, a voter, a consumer. Participate deliberately.