🏛 Global Politics × AI · Linh Truong
🏛 Personal notes · May 2026

Global Political Transformation in the AI Era

I built this note to track how AI is reshaping power, governance, and legitimacy worldwide — across regimes, elections, the information ecosystem, the state itself, and who holds authority. It is my working map of concentration vs. accountability, not a forecast of which wins.

The question behind this note: will cheap intelligence strengthen accountability — or concentrate surveillance, persuasion, and executive power faster than institutions can check it? What follows is my read of the forces, the frictions, and the levers that actually matter for democracies, civil society, and the multilateral order.
📍 Scope: Power · Governance · Legitimacy 📅 Horizon: 2022 → 2035 🧭 Stance: Accountability is a choice ✍️ By: Linh Truong
~50%
Of humanity lived in countries holding national elections in 2024 — the largest election year ever, and AI's first global test
19 yrs
Consecutive years of decline in global freedom (Freedom House); V-Dem finds most people now live under autocratizing regimes
3 models
Rival governance approaches crystallizing — EU rights-based, US market-led, China state-control — each exporting its template
2 paths
AI can entrench surveillance autocracy or renew democratic capacity — institutions, not the technology, decide which
01 · What I'm tracking

The political story in one page

AI is a power technology — it cheaply scales the core functions of governance: surveillance, persuasion, prediction, and administration. Whoever wields it can amplify control or accountability. To me, the impact turns less on frontier capability than on who controls the tools and whether institutions can check their use. The question I keep weighing: concentration or accountability?

Mechanism

AI changes the cost of control

It collapses the cost of watching, persuading, and deciding at scale. Functions that once required armies of officials now run on models — shifting the balance between rulers and ruled.

Stakes

Who holds power

AI reshapes the relationship between states and citizens, executives and institutions, and great powers. Compute, data, and platforms become instruments of statecraft.

Battleground

The information commons

Synthetic media and algorithmic curation strain the shared reality democracies depend on — while giving autocracies new tools for narrative control.

🎯

What I keep coming back to: Politics faces a fork. One path is concentration — AI becomes a force-multiplier for surveillance, censorship, and unchecked executive power, hardening a durable "surveillance authoritarianism." The other is renewed accountability — AI strengthens public administration, citizen voice, oversight, and the rule of law. Which path wins is an institutional and political choice, not a technological inevitability. The same model that powers a censorship engine can power an ombudsman.

02 · The Big Picture

The political AI power system — how capability becomes authority

A small set of capability inputs (compute, models, data, platforms) flow through the mediating institutions of politics — the state, the information ecosystem, elections, and civil society — to produce political outcomes: how power is distributed, whether rule is legitimate, and who holds rights. Feedback loops either entrench control or reinforce accountability.

FIG 1

Capability inputs → political institutions → power outcomes → feedback

1 · CAPABILITY INPUTS 2 · POLITICAL INSTITUTIONS 3 · POWER OUTCOMES 4 · POLITICAL ORDER Compute & Models frontier & open systems Data & Surveillance biometrics · records · CCTV Platforms & Reach social · search · messaging Capital & Talent who funds & builds it The State bureaucracy · courts · police Information Ecosystem media · public sphere Elections & Parties campaigns · representation Civil Society press · NGOs · oversight Distribution of Power concentrated vs. checked Legitimacy & Trust consent · shared reality Rights & Liberties privacy · speech · due process Regime Type democracy ↔ autocracy State Capacity effective vs. predatory Geopolitical Power influence & alliances Social Stability cohesion vs. unrest + Accountability loop: oversight, transparency, contestable elections − Control loop: surveillance, censorship, executive capture ⚖ Governance gateway law · checks · norms
Read it left→right: the same capability inputs flow through political institutions, but the governance gateway — law, checks, and norms — determines whether they produce concentrated or accountable power. The green loop = accountability that disperses power; the red loop = control that concentrates it. The technology is identical; the wiring differs.
03 · The World Map

Regime landscape — who is transforming, and how

Different political systems absorb AI very differently. The technology is a regime amplifier: it sharpens whatever logic already governs a state. This is the heart of the "global" story — the transformation is many transformations, pulling in opposite directions.

TBL 1

How regime types use — and are reshaped by — AI

Regime archetypeState AI capacityChecks & accountabilityPrimary use of AIDominant risk
Closed autocracies
(e.g. China model)
Very highMinimalSurveillance, censorship, social management, "AI+" governanceDurable techno-authoritarian consolidation
Electoral autocracies
(competitive authoritarian)
Medium–highWeakInformation control, targeted repression, manufactured consentTilting the playing field; sham legitimacy
Backsliding democraciesMediumErodingExecutive aggrandizement, partisan capture of toolsDemocratic erosion accelerated by AI
Liberal democracies
(rights-based)
MediumStrong (contested)Public-service delivery, regulation, contested speech debatesTrust collapse; capacity & legitimacy gaps
Weak / fragile statesLowVariablePatchy adoption; reliance on external platformsForeign info-ops; capture; instability
Multilateral / supranational
(EU, UN bodies)
Rule-makerInstitutionalStandard-setting, rights frameworks, coordinationSlow, fragmented, enforcement-limited
Legend: Strong/Low Medium/Rule-maker Medium/Eroding Weak/Minimal Very high. Archetypes are directional, not country rankings — most real states are hybrids and move between rows as AI capacity and institutional health shift.
FIG 2

State AI capacity vs. democratic accountability — where regimes sit

State AI capacity → Accountability & checks on power → SURVEILLANCE STATE high capacity · low accountability CAPABLE DEMOCRACY high capacity · high accountability FRAGILE / CAPTURED OPEN BUT LOW-CAPACITY China model Electoral autocracies Backsliding dem. EU / Nordics US (contested) UK · Japan India Small open dem. Fragile states Positions are directional. AI tends to push states upward (more capacity) — the open question is whether accountability keeps pace (rightward) or not.
The danger zone is the top-left: rising state AI capacity with weak checks (the surveillance state). The goal is the top-right: capacity matched by accountability (the capable democracy). AI is steadily pushing every state upward; the strategic task is to push rightward at the same time.
04 · The Concentration Problem

How AI concentrates political power

AI lowers the cost of three things rulers have always wanted: seeing everything, shaping what people believe, and acting without friction. When these become cheap and automatable, the historical limits on centralized power weaken — the core force pushing toward authoritarian consolidation.

FIG 3

The control toolkit — what AI makes cheap (illustrative)

SEE PERSUADE ACT Mass surveillance: face/voice ID, location, behavior prediction at scale human-limited before Mass persuasion: targeted messaging, synthetic media, bot networks cost of producing content → ~0 Automated decisions: benefits, policing, scoring, agentic enforcement friction & accountability → fall Bars depict the collapsing cost/limit of each function, not measured shares — the political point is that old natural limits are disappearing.
Historically, the cost of watching, persuading, and administering at scale capped how much power could centralize. AI removes those caps. The same capabilities, under strong checks, instead deliver better public services and oversight — which is precisely why the governance gateway matters more than the technology.
See

Surveillance at scale

Biometric ID, CCTV analytics, and data fusion make pervasive monitoring affordable. Predictive policing and social scoring turn observation into pre-emptive control.

Persuade

Industrial-scale influence

Synthetic media and micro-targeting let a single actor flood the public sphere — manufacturing consent or drowning dissent at near-zero marginal cost.

Act

Automated statecraft

Agentic systems execute decisions — approvals, denials, enforcement — faster than humans can review, eroding due process and the friction that protects rights.

Executive drift

Power slips the leash

AI lets leaders bypass bureaucracies, courts, and parties — concentrating capability in the executive and weakening internal checks.

Asymmetry

The watcher's advantage

Those who hold the data and models gain a structural edge over those they govern — a knowledge asymmetry that compounds over time.

Countervailing

Tools cut both ways

Open models, leak analysis, and citizen tech also empower watchdogs, journalists, and courts — the strongest force against concentration.

05 · The Epistemic Battleground

The information ecosystem & shared reality

Democracy runs on a common factual baseline and the ability to trust what you see. AI strains both — not mainly through any single convincing fake, but by making synthetic content abundant, cheap, and ambient, which corrodes trust in everything.

The threat

Pollution of the commons

  • Synthetic media (deepfakes, cloned voices, AI imagery) blurs the line between real and fabricated.
  • The "liar's dividend": when anything could be fake, the powerful dismiss real evidence as fabricated.
  • Scale & speed: bot networks and AI content farms flood feeds faster than verification can keep up.
  • Algorithmic curation sorts citizens into divergent realities, rewarding outrage over accuracy.
The reality check

Resilience & nuance

  • 2024's verdict: the feared "deepfake election apocalypse" was more limited than predicted — but cheap fakes and erosion of trust did real damage in specific cases.
  • Detection & provenance (content credentials, watermarking, C2PA) offer partial defenses.
  • Institutional trust — strong independent media, fact-checking, and electoral bodies — is the real firewall.
  • AI also aids defenders: rapid debunking, translation, and open-source investigation.
🪞

The epistemic paradox: the deepest danger is not that people believe a specific fake — it is that they stop believing anything. A world where all evidence is contestable favors whoever controls attention and narrative, which structurally advantages the powerful and the loudest. Defending shared reality is now a core function of democratic governance.

06 · The Trajectory

Four phases — and the diverging political paths

As AI capability diffuses, regimes split. Autocracies adopt the control toolkit quickly and decisively; democracies move slower, constrained (and protected) by checks. The gap in trajectory widens in the middle phases before the long-run outcome is settled by institutional response.

FIG 4

Diverging paths — power concentration over time, by regime

Power concentration → Time → 2022–242024–27 2027–302030–35 Autocracies (control toolkit) Backsliding democracies Resilient democracies (checks hold) ⟵ widening authoritarian advantage ~today's baseline
Autocracies capture the control toolkit early and concentrate power fast. Whether democracies hold the line (green) or drift upward (amber) depends on whether oversight, courts, press, and elections adapt as quickly as the technology does. The middle phases are the most dangerous window.
PHASE 1 · 2022–24

Disruption & alarm

Generative AI goes mainstream; 2024's record election year becomes the first global stress test. Rules lag the technology.

PHASE 2 · 2024–27

Institutionalization

States build AI into governance. EU AI Act phases in; the US pivots to a deregulatory, pro-deployment posture; China deepens "AI+" control. Blocs harden.

PHASE 3 · 2027–30

Agentic governance

Autonomous systems make and execute decisions. The fight over due process, oversight, and the "human in the loop" intensifies.

PHASE 4 · 2030–35

New equilibrium

A more bloc-structured world order settles. Either accountable AI-augmented governance or durable surveillance autocracy hardens — path-dependent.

07 · The Machinery of Government

AI and the state itself

Beyond elections and rights, AI is rewiring the day-to-day operation of government — how services are delivered, decisions are made, and the bureaucracy functions. This is where the largest, quietest transformation is happening, for better and worse.

Upside

The capable state

AI can cut backlogs, translate services into every language, detect fraud, model policy, and bring responsive government to citizens long underserved — a genuine accountability dividend if done transparently.

Downside

The opaque state

Automated decision-making can deny benefits, flag citizens, or allocate enforcement with no explanation and no appeal — embedding bias and removing the human judgment due process requires.

Shift

The hollowed state

As capability moves into models built by a few firms, states risk dependence on private vendors for core functions — outsourcing sovereignty and creating new chokepoints.

FIG 5

Where AI enters the machinery of government

Policy designsimulation · drafting Administrationbenefits · permits · tax Service deliveryhealth · education · 311 Coercionpolicing · borders · justice Oversightaudit · courts · appeal + efficiency ⚠ due process + access ⚠ rights risk ⚖ must keep pace The further right AI penetrates without matching oversight, the higher the rights & legitimacy risk. Healthy adoption strengthens the oversight box as fast as the coercion box — most failures don't.
AI is most benign in service delivery and most dangerous in coercion. The decisive variable is whether the oversight function (audit, courts, the right to appeal, explainability) is upgraded as aggressively as the front-line tools. When it lags, automation quietly strips accountability.
08 · Elections & Voice

Democracy, elections & citizen participation

2024 put democracy under the AI microscope: more than 60 countries and roughly half of humanity went to the polls. The result was neither catastrophe nor non-event — a nuanced picture that defines the real risks and the real defenses.

Pressure points

How AI strains elections

  • Synthetic campaign content: cloned candidate voices, fake endorsements, fabricated scandals (e.g. robocalls, manipulated audio).
  • Micro-targeted manipulation: tailored, often misleading messaging at individual scale.
  • Foreign interference: state-backed influence operations, now cheaper and harder to attribute.
  • Voter suppression & confusion: fake logistics info, AI-generated doubt about results.
Democratic upgrades

How AI can strengthen democracy

  • Deliberative tools: AI-facilitated citizens' assemblies and large-scale consensus-finding (e.g. pol.is-style sensemaking).
  • Constituent service: faster, multilingual access to representatives and government.
  • Transparency: parsing budgets, bills, and lobbying records for public scrutiny.
  • Turnout & access: personalized, accurate civic information.
FIG 6

The 2024–25 election stress test — feared vs. observed

Impact ↑ Decisive fakes Trust erosion Cheap fakes / volume Foreign ops swinging results Feared (pre-2024) Observed (2024–25)
The headline lesson: the apocalyptic scenarios (a single deepfake swinging a major election) largely did not materialize, but trust erosion, the "liar's dividend," and high-volume cheap fakes were real and persistent. The defense that worked was institutional: resilient media, prepared electoral bodies, and rapid debunking — not detection technology alone.
09 · The Governance Contest

Three models, one fragmenting order

There is no single global government of AI. Instead, three regulatory models are crystallizing and competing to set the world's rules — each exporting its template through trade, standards, and infrastructure. This is the geopolitics of AI governance.

FIG 7

The three governance models — and the coordination gap

🇪🇺 RIGHTS-BASED EU / "Brussels effect" Risk-tiered law (AI Act), fundamental rights first, prohibited uses, audits. Exports rules via market access 🇺🇸 MARKET-LED US / "Washington effect" Pro-innovation & deployment, light federal rules, security focus, export controls. Exports capability & standards 🇨🇳 STATE-CONTROL China / "Beijing effect" Content & algorithm control, "AI+" state agenda, labeling & security mandates. Exports infrastructure & model ⚠ The coordination gap: no binding global regime — only soft processes (UN panel, summits, OECD principles) Most of the world's states must choose, blend, or be caught between these templates.
Whoever sets the standards shapes global markets and norms — the "regulatory power" contest. For most countries the real choice is which template to adopt or how to blend them. The thin layer of genuine multilateralism (UN AI processes, the safety-summit series, OECD principles) coordinates norms but cannot yet bind.
EU AI Act

Rights as default

The first comprehensive AI law, phasing in 2025–2027: bans on certain uses, strict rules for high-risk systems, transparency for general-purpose models.

US pivot

Deregulatory turn

The 2025 shift rescinded the prior safety-focused executive order for a pro-deployment, competitiveness-and-security agenda — loosening federal constraints.

China's "AI+"

State-directed

Deep-synthesis labeling, algorithm registries, and security reviews fuse AI to state objectives — and are exported with infrastructure.

Global layer

Norms, not binding

UN scientific panel & global dialogue, the Bletchley→Seoul→Paris summit arc, OECD/UNESCO principles — convening power without enforcement.

10 · Citizens & Liberty

Rights, surveillance & the citizen

The clearest political effect of AI is on the relationship between the individual and the state. Privacy, free expression, due process, and freedom from arbitrary power are all directly in play — and the spectrum of outcomes is wide.

FIG 8

The surveillance spectrum — from privacy to total visibility

Strong privacydata rights, limits Targeted, warrantedoversight & due process Pervasive monitoringnormalized, weak limits Predictive controlscoring, pre-emption AI lowers the cost of moving rightward. Holding a state on the left half is now an active, ongoing political fight.
Every state sits somewhere on this spectrum, and AI makes drifting rightward cheaper and more tempting. Where it lands is set by law (data protection, warrant requirements), independent courts, and a free press — the institutional brakes that AI does not weaken on its own.
Privacy

The end of obscurity

Data fusion and biometric ID erode the practical anonymity that protected dissent. "Nothing to hide" collapses when everything is recorded and inferable.

Expression

Chilling & censorship

Automated content moderation and the knowledge of being watched both suppress speech — sometimes by design, sometimes as a side effect.

Due process

The unaccountable decision

Algorithmic scoring in policing, welfare, and justice can deny rights with no explanation, no appeal, and embedded bias.

Assembly

Organizing under watch

Protest and association become riskier when participants are auto-identified — reshaping the dynamics of dissent and reform.

Empowerment

Tools for citizens

The same AI helps citizens know their rights, document abuses, navigate bureaucracy, and hold power to account.

Equity

Whose rights, unevenly

Harms fall hardest on the marginalized and the surveilled; benefits accrue to the connected. AI can widen or narrow civic inequality.

11 · The Strategy

Playbooks for an accountable AI politics

Different actors hold different levers. I use this stack to check whether multilateral norms, national law, and civil society are keeping accountability ahead of control — or letting capacity outrun checks.

FIG 9

The governance strategy stack

MULTILATERAL shared norms, safety & rights floors NATIONAL / STATE law, oversight, capable & checked government CIVIL SOCIETY & CITIZENS press, courts, watchdogs, an informed public coordinate ↑ legislate & oversee → contest & verify →
"Renewed Democracy" requires all three layers together: a multilateral floor on rights and safety, national law and oversight that keep pace with deployment, and a vigorous civil society that can contest and verify. Remove any layer and the system tilts toward concentration.
For Democratic Governments

Build capacity, keep checks ahead of it

  • Upgrade oversight first: fund audit, courts, and appeal rights to move before front-line automation.
  • Legislate guardrails: rules for government AI use, bans on the most dangerous applications, transparency mandates.
  • Protect elections: provenance standards, rapid-response units, electoral-body preparedness.
  • Defend privacy: strong data protection and limits on biometric mass surveillance.
  • Stay sovereign: avoid over-dependence on single vendors for core state functions.
For Civil Society & Citizens

Contest, verify, and demand accountability

  • Use AI for oversight: investigative tooling, document analysis, abuse documentation.
  • Build epistemic resilience: media literacy, provenance habits, trusted verification.
  • Demand transparency: the right to know when AI decides, and to appeal.
  • Protect the watchdogs: independent press, courts, and election monitors.
  • Organize: coalitions that keep rights on the political agenda.
For Multilateral Bodies

Set floors, coordinate, narrow the gap

  • Establish rights & safety floors: shared red lines (e.g. autonomous coercion, mass surveillance norms).
  • Coordinate standards: interoperable provenance, evaluation, and transparency regimes.
  • Support fragile states: capacity-building so they aren't captured by foreign templates or info-ops.
  • Convene credibly: sustain the safety-summit and UN-panel processes toward binding commitments.
  • Manage cross-border risk: election interference, info-ops, and proliferation.
For Platforms & Builders

Build for accountability by default

  • Provenance & labeling: content credentials and watermarking as standard.
  • Election integrity: policies, rapid response, and refusal of clearly harmful political uses.
  • Transparency: explainability and contestability for consequential decisions.
  • Resist misuse: safeguards against surveillance and influence-operation tooling.
  • Empower defenders: tools and access for journalists, researchers, and oversight bodies.
12 · What Could Go Wrong

Political risk & governance map

The political risks of AI cross borders and institutions, making them hard for any single actor to manage. Mapping them by likelihood and severity clarifies where to act first.

FIG 10

Political risk quadrant — likelihood × severity

Severity → Likelihood (near-term) → monitor manage actively watch tail risks act now Surveillanceautocracy Trust /epistemic collapse Democraticerosion Electioninterference Algorithmicinjustice Blocfragmentation Autonomouscoercion Vendorcapture
The signature political risks cluster in "act now": surveillance autocracy, epistemic/trust collapse, accelerated democratic erosion, and election interference. Tail risks — autonomous coercion, governance fragmentation — are lower-likelihood but severe enough to demand coordinated attention precisely because no state can contain them alone.
Concentration

Surveillance autocracy

AI hardens durable authoritarian control — pervasive monitoring, automated repression, and manufactured consent that resists reform.

Erosion

Democratic backsliding

Even in democracies, executives capture AI tools, weaken checks, and tilt the playing field — slow drift rather than sudden coup.

Epistemics

Trust collapse

A polluted information commons where nothing is believed corrodes the shared reality democracy and accountability require.

Fragmentation

Governance splinternet

Rival regulatory blocs and a "splinternet" reduce cooperation on shared risks and force states to pick sides.

Stability

Unrest & legitimacy

Rapid, opaque automation of state functions without consent risks backlash, polarization, and instability.

Response

The governance toolkit

Rights law, oversight upgrades, provenance standards, election defense, multilateral floors, and a protected civil society.

13 · Where This Goes

Political scenarios, 2030–2035

Two axes define the political outcome space: how power is distributed (concentrated vs. dispersed) and how accountable institutions are (unchecked vs. accountable). The world will not land in one quadrant uniformly — but every state's trajectory bends toward one of these four.

FIG 11

Scenario matrix — power distribution × accountability

ACCOUNTABLE institutions UNCHECKED institutions CONCENTRATED power DISPERSED power 🏛 Technocratic Guardrails Power stays concentrated but is checked by law, courts, and transparency. Capable, orderly — but elite-driven & paternalistic. Stable · narrow 🌍 Renewed Democracy Dispersed power + strong accountability. AI augments oversight, participation, and capable, trusted government. Renewed Democracy · north star 👁 Surveillance Leviathan Concentrated, unchecked power. Pervasive monitoring, automated repression, durable autocracy. The most dangerous quadrant. Stable · oppressive 🌀 Chaotic Fragmentation Power disperses but accountability fails. Eroded trust, info chaos, weak states, capture by whoever shouts loudest. Dynamic · unstable
The target is the top-right — Renewed Democracy: power dispersed and institutions accountable. The most dangerous is the bottom-left — the Surveillance Leviathan. Strategy (§11) is the work of moving states up and to the right: keeping accountability ahead of capacity, and dispersing rather than concentrating control.
14 · Track Progress

Political metrics that matter

Signals to watch to know whether the world is bending toward accountability or concentration — and how individual states are positioned.

Global / System level

Accountability vs. concentration signals

Democracy indicesV-Dem & Freedom House trajectories — net gains or declines
Surveillance spreadAdoption of biometric mass-surveillance systems by states
Information integrityTrust in media; documented influence operations
Governance convergenceBinding multilateral commitments vs. bloc fragmentation
Election integrityIncidents, resilience, and confidence across major votes
Provenance adoptionShare of content carrying credible content credentials
Rights red linesBans on the most dangerous uses holding or eroding
National level

State health dashboard

Checks-vs-capacity gapIs oversight keeping pace with state AI capability?
Transparency rightsRight to know & appeal automated decisions
Privacy protectionData-protection strength; limits on biometrics
Judicial independenceCourts able to review state AI use
Press & civil societyFreedom and capacity of watchdogs
Vendor dependenceReliance on single providers for core functions
Public AI literacyCitizen ability to navigate the information ecosystem
⚠️

Watch the leading indicator: the gap between a state's AI capacity and the strength of its checks and accountability is the single best early signal of which scenario it is heading toward. A widening gap = drift toward the Surveillance Leviathan; a closing gap = movement toward Renewed Democracy.

15 · Remember This

Ten things I keep on my desk

Not commandments — reminders I re-read when the political noise gets loud.

1 · AI is a power technology

It cheaply scales surveillance, persuasion, prediction, and administration — the core functions of governance.

2 · Concentration vs. accountability is THE question

The same model can run a censorship engine or an ombudsman. Institutions, not the technology, decide.

3 · AI is a regime amplifier

It sharpens whatever logic already governs a state — entrenching autocracies and stress-testing democracies.

4 · The real fight is keeping checks ahead of capacity

AI pushes every state toward more capability; accountability must move at least as fast.

5 · The deepest info threat is disbelief, not deception

The "liar's dividend" — where nothing is trusted — favors the powerful and the loudest.

6 · 2024 was a stress test, not an apocalypse

Feared deepfake catastrophes were limited; institutional resilience, not detection tech, was the firewall.

7 · Three governance models are competing

EU rights-based, US market-led, China state-control — each exporting its template into a coordination gap.

8 · The quiet revolution is inside the state

Automated administration and coercion reshape governance daily — oversight must be upgraded alongside.

9 · Tail risks need international coordination

Election interference, autonomous coercion, and info-ops cross borders; no state can contain them alone.

10 · Aim for Renewed Democracy

Dispersed power + accountable institutions is the only quadrant that is both free and stable.

16 · References & Sources

Where the ideas in this note come from

Annotated bibliography behind the concentration-vs-accountability thesis, political power-system map, regime table and quadrant, control-toolkit diagram, epistemic battleground, diverging trajectories, state machinery, 2024 election stress test, three governance models, surveillance spectrum, strategy stack, risk quadrant, scenario matrix, and KPI dashboard. Section tags (e.g. §05) show where each source is used. Diagrams and operating rhythms are my synthesis unless noted.

Scope. Synthesis of democracy research, digital-authoritarianism studies, election-integrity reporting, and AI-governance sources (May 2026). Hero-strip figures (e.g. ~50% of humanity voting in 2024, 19 consecutive years of global freedom decline, three rival governance models) blend IDEA, Freedom House, V-Dem, and official regulatory documents — directional claims, not forecasts for any country or election. Not legal, policy, or political advice.

Citations are numbered continuously [1]–[n] within this section.

Democracy recession, elections & the global regime landscape (§01, §03, §08, hero, §14)

  1. International IDEA, Global Overview of Elections 2024. 2024. More than 60 countries holding national elections; roughly half of humanity eligible to vote — hero “~50%” stat and §08 “2024 stress test” framing. idea.int — hero, §08.
  2. Freedom House, Freedom in the World 2025. 2025. 19th consecutive year of global freedom decline; country scores and autocratization trends — hero stat and §14 democracy-index KPIs. freedomhouse.org — hero, §03, §14.
  3. Freedom House, Freedom on the Net (annual). Digital rights, censorship, and surveillance online — §05 information commons and §10 expression cards. freedomhouse.org/freedom-net — §05, §10.
  4. V-Dem Institute, Democracy Report 2025. 2025. Autocratization affecting a record share of the world population; regime-type trajectories — hero “autocratizing regimes” line, TBL 1 archetypes, FIG 2 positioning. v-dem.net — hero, §03, §14.
  5. Economist Intelligence Unit, Democracy Index (annual). Global democracy scores and “democratic recession” narrative — §01 fork and §14 indices row. eiu.com/democracy-index — §01, §14.
  6. Levitsky, S. & Ziblatt, D., How Democracies Die. Crown, 2018. Executive aggrandizement, norm erosion, and competitive erosion of checks — §03 backsliding row and §12 democratic-erosion bubble. — §03, §12.
  7. Levitsky, S. & Way, L. A., Competitive Authoritarianism: Hybrid Regimes after the Cold War. Cambridge University Press, 2010. Electoral autocracy archetype in TBL 1 and FIG 2 “electoral autocracies” label. — §03, FIG 2.

AI as power technology & the political system map (§01–§02, FIG 1)

  1. Kissinger, H., Schmidt, E., & Huttenlocher, D., The Age of AI: And Our Human Future. Little, Brown, 2021. AI reshaping statecraft, legitimacy, and the balance of power — §01 “power technology” cards and §02 capability→authority flow. — §01, §02.
  2. Schneier, B., “The Impact of Artificial Intelligence on Democracy.” Testimony and essays, 2023–25. AI scaling surveillance, persuasion, and administrative control — §01 mechanism cards and §04 control toolkit. schneier.com — §01, §04.
  3. AI Now Institute, AI Now 2023 Landscape: Confronting Tech Power and subsequent reports. Concentration of AI capability in firms and states; governance gaps — §02 inputs layer and §04 executive-drift card. ainowinstitute.org — §02, §04, §12.
  4. Stanford HAI, 2025 AI Index Report — Policy & Governance chapter. 2025. National AI strategies, regulation, and adoption by regime type — TBL 1 capacity column and §06 Phase 2 institutionalization. hai.stanford.edu/ai-index — §03, §06, §09.

Digital authoritarianism, surveillance & power concentration (§04, §10, FIG 3, FIG 8)

  1. Feldstein, S., The Rise of Digital Repression: How Technology is Reshaping Power, Politics, and Resistance. Oxford University Press, 2021. AI-enabled surveillance, predictive policing, and digital repression — §04 SEE/PERSUADE/ACT cards and FIG 8 spectrum. — §04, §10, FIG 8.
  2. Carnegie Endowment for International Peace, AI Global Surveillance Index (Feldstein et al.). 2019–23 updates. State adoption of AI surveillance by country — §03 closed-autocracy row and §14 surveillance-spread KPI. carnegieendowment.org — §03, §14.
  3. Zuboff, S., The Age of Surveillance Capitalism. PublicAffairs, 2019. Behavioral data extraction and asymmetric knowledge — §04 “watcher’s advantage” and §02 data/surveillance input node. — §02, §04.
  4. Access Now, Surveillance Tech & Human Rights reports and #KeepItOn campaign. Biometric mass surveillance and shutdown trends — §10 privacy/assembly cards and §11 privacy bullets. accessnow.org — §10, §11.
  5. Human Rights Watch & Amnesty International, China social-credit and mass-surveillance reporting. 2019–25. Pervasive monitoring and automated social management — §03 “China model” in TBL 1/FIG 2 and §04 surveillance-at-scale card. hrw.org — §03, §04.

Information integrity, synthetic media & the epistemic commons (§05, §08, FIG 6)

  1. Chesney, R. & Citron, D. K., “Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security.” California Law Review, 2019. Synthetic media risks and the “liar’s dividend” — §05 threat cards and takeaway #5. californialawreview.org — §05, §15.
  2. Donovan, W. & Friedberg, B., “The Source Hacking Media Manipulation Tactics.” Data & Society, 2019. Cheap fakes, volume, and trust erosion — §05 “pollution of the commons” and FIG 6 cheap-fakes bar. datasociety.net — §05, FIG 6.
  3. Stanford Internet Observatory, election integrity & influence-operation research (2024 cycle). Documented cases, platform responses, and limits of decisive fakes — §08 pressure points and FIG 6 “feared vs. observed.” stanford.io — §05, §08, FIG 6.
  4. Alan Turing Institute (CETaS), AI and the 2024 Elections: An Overview. 2024. Global election-year AI incidents and institutional defenses — §08 “2024 verdict” card and FIG 6 caption. turing.ac.uk — §08, FIG 6.
  5. Oxford Internet Institute, Computational Propaganda Project / Industrialized Disinformation reports. State and commercial influence operations at scale — §08 foreign-interference bullet and §12 election-interference bubble. comprop.oii.ox.ac.uk — §08, §12.
  6. Reuters Institute for the Study of Journalism, Digital News Report (annual). Trust in news, platform news use, and information diets — §05 epistemic paradox and §14 information-integrity KPI. reutersinstitute.politics.ox.ac.uk — §05, §14.
  7. Coalition for Content Provenance and Authenticity (C2PA), content-credentials standard. 2022–25. Provenance and watermarking for synthetic media — §05 detection/provenance defenses and §11 platform bullets. c2pa.org — §05, §11, §14.

Diverging trajectories, the state & algorithmic governance (§06–§07, FIG 4–5)

  1. European Parliament & Council, Regulation (EU) 2024/1689 (AI Act). 2024. Phased implementation 2025–2027; high-risk and prohibited uses — §06 Phase 2 and §09 EU card. eur-lex.europa.eu — §06, §09.
  2. White House, Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence (EO 14110). 30 October 2023. Prior US safety-and-rights framework superseded by the 2025 pro-deployment pivot referenced in §09 — baseline for §06 Phase 2 contrast. whitehouse.gov — §06, §09.
  3. White House, AI Action Plan and related 2025 federal AI policy documents. Pro-innovation, deregulatory, and competitiveness posture — §09 “US pivot” card and §06 bloc-hardening phase. whitehouse.gov (verify latest releases). — §06, §09.
  4. Cyberspace Administration of China, Provisions on the Administration of Deep Synthesis Internet Information Services and generative-AI measures. 2022–23. Labeling, registry, and security reviews — §09 China “AI+” card and §05 synthetic-media control. cac.gov.cn — §05, §09.
  5. Eubanks, V., Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor. St. Martin’s Press, 2018. Automated welfare and enforcement without due process — §07 opaque-state card and FIG 5 coercion box. — §07, §10.
  6. O’Neil, C., Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy. Crown, 2016. Opaque scoring and accountability failures — §07 administration/coercion risks and §10 due-process card. — §07, §10, §12.
  7. Partnership on AI & OECD, public-sector AI adoption and accountability guides. 2023–25. Government use cases, human oversight, and vendor dependence — §07 hollowed-state card and §11 sovereignty bullets. partnershiponai.org — §07, §11.

Three governance models & regulatory competition (§09, FIG 7)

  1. Bradford, A., The Brussels Effect: How the European Union Rules the World. Oxford University Press, 2020. EU regulatory export via market access — §09 “Brussels effect” and FIG 7 rights-based model. — §09, FIG 7.
  2. OECD, OECD AI Principles. 2019 (updated 2024). Interoperable trust, accountability, and human-centered values — §09 global-layer card and §11 multilateral floor. oecd.org/ai-principles — §09, §11.
  3. UNESCO, Recommendation on the Ethics of Artificial Intelligence. 2021. Global ethics framework adopted by 193 member states — §09 norms layer and §10 rights framing. unesco.org — §09, §10.
  4. Centre for the Governance of AI (GovAI), State of AI Policy tracking. Comparative national AI governance models — FIG 7 three-bloc diagram and hero “3 models” chip. governance.ai — hero, §09.
  5. Brookings Institution, AI governance & “Washington effect” analyses. US standards export through platforms and capability — §09 market-led model description. brookings.edu — §09.

Global coordination, safety summits & multilateral processes (§09–§11, §14)

  1. United Nations, Governing AI for Humanity — High-level Advisory Body final report. September 2024. Global governance gaps and proposed scientific panel — §09 coordination-gap line and §11 multilateral layer. un.org/governing-ai-for-humanity — §09, §11.
  2. United Nations, Independent International Scientific Panel on AI & Global Dialogue on AI Governance. 2025 launch. Follow-on to HLAB recommendations — §09 global-layer card and §14 governance-convergence KPI. un.org/ai-advisory-body — §09, §14.
  3. UK Government, Bletchley Declaration — AI Safety Summit, 1–2 November 2023. First major multilateral safety communiqué — §09 summit arc. gov.uk — §09.
  4. Republic of Korea & UK, Seoul Declaration for Safe, Innovative and Inclusive AI. AI Seoul Summit, 21 May 2024. Frontier-risk commitments — §09 summit arc. gov.uk — §09.
  5. France & India, AI Action Summit — Paris, 10–11 February 2025. Shift toward action and competitiveness; partial US/UK abstentions — §09 summit arc and §12 bloc-fragmentation risk. elysee.fr — §09, §12.
  6. International AI Safety Report 2025 (Y. Bengio, chair). January 2025. Shared scientific assessment across governments — §11 multilateral “safety floors” and §12 tail-risk framing. internationalaisafetyreport.org — §11, §12.

Rights, privacy & algorithmic accountability (§10–§11, §14)

  1. Regulation (EU) 2016/679 (General Data Protection Regulation). 2016. Data rights, purpose limitation, and automated-decision safeguards — §10 privacy card and §14 privacy-protection KPI. eur-lex.europa.eu — §10, §14.
  2. Algorithmic Justice League & academic algorithmic-bias literature (Buolamwini, Gebru et al.). Documented bias in facial recognition and automated systems — §10 equity card and §12 algorithmic-injustice bubble. ajl.org — §10, §12.
  3. NIST, Artificial Intelligence Risk Management Framework (AI RMF 1.0). 2023. Risk-based governance template for consequential AI — §11 guardrails and §14 transparency-rights KPI. nist.gov/ai-rmf — §11, §14.
  4. Farrell, H. & Newman, A. L., “Weaponized Interdependence: How Global Economic Networks Shape State Coercion.” International Security, 2019. Platform and infrastructure chokepoints — §02 platform input and §12 vendor-capture bubble. doi.org/10.1162/isec_a_00351 — §02, §12.

Democratic renewal, deliberation & civil-society leverage (§08, §11, §13)

  1. Collective Intelligence Project & pol.is / Participatory Democracy Foundation. Large-scale sensemaking and AI-assisted deliberation — §08 deliberative-tools bullet and §11 civil-society stack. pol.is — §08, §11.
  2. DeepMind, “AI-assisted deliberation” research (e.g. Habermas Machine / opinion synthesis experiments). 2024–25. AI facilitating consensus-finding in citizens’ assemblies — §08 democratic-upgrades card. deepmind.google (search “deliberation”). — §08.
  3. Larry Diamond, “Democratic Recession” essays and Ill Winds. Hoover Institution / Penguin, 2019. Global democratic decline and renewal levers — §01 accountability fork and §13 Renewed Democracy north star. hoover.org — §01, §13.
  4. Transparency International, Corruption Perceptions Index and open-government data initiatives. Oversight capacity and state capture — §11 watchdog protection bullets and §14 press/civil-society KPI. transparency.org — §11, §14.

Author synthesis & companion notes

  1. Truong, L., Global Political Transformation in the AI Era — personal working notes. May 2026. Original diagrams (FIG 1–11, TBL 1), governance strategy stack, scenario matrix, risk quadrant, and ten desk principles. LinhTruong.com — all sections.
  2. Truong, L., companion notes: Global Economic Transformation in the AI Era and Geopolitics Transformation in the AI Era. Cross-border economic diffusion, compute geopolitics, and bloc competition frames referenced in §02 feedback loops, §09 governance contest, and §12 fragmentation scenario. Same author collection. — §02, §09, §12.
📎

Before you quote externally: FIG 3 control-toolkit bars and FIG 6 election-impact bars are illustrative — they depict collapsing costs and feared-vs-observed ranges, not measured shares. The hero “~50% of humanity” figure compresses IDEA’s 2024 election-year estimate; Freedom House’s “19 years” and V-Dem’s autocratization share use different methodologies — do not merge them into a single index. US AI policy shifted materially in early 2025; verify current executive orders and agency rules before citing. Country positions in FIG 2 are directional snapshots, not rankings.

↑ Top