Every page.
A clear way through.

Explore Mutual AI Resilience, civilian continuity and policy research. Examine evidence, compare scenarios and assess AI-assisted cooperation.

A detailed explanation of MAR: sovereign participation, verified civilian warnings, the role of AI, operating responsibilities, safeguards, limits, and evaluation.

Explore 20 Mutual AI Resilience hypotheses, compare assumptions, run fictional civilian outcome estimates, and learn how to test a claim about cooperation.

How Safe Nations separates normative policy, hypotheses, context, and findings; designs comparisons; measures harm; and keeps research reproducible.

Answers about MAR, sovereignty, civil liberties, AI value, policy commitments, costs, evidence, the first 100 days, the lab, and political independence.

Eight proposed commitments for sovereign AI resilience, civilian service continuity, civil liberties, practical diplomacy, accountable spending, and oversight.

A staged proposal for evaluating Mutual AI Resilience, with responsibilities, deliverables, decision gates, stop conditions, and an illustrative first 100 days.

Proposed responsibilities, acceptance tests, suspension triggers, information controls, independent review, and public accountability for Mutual AI Resilience.

Proposed MAR applications in trade, oil, diesel, energy, water, food, health, communications, and diplomacy, with sector-specific questions and linked research records.

The principles informing Safe Nations: sovereign authority, practical cooperation, civil liberties, and accountable institutions, with selected dated public sources.

A print-friendly executive brief on MAR: the public purpose, proposed commitments, operating model, evidence limits, first decisions, and evaluation plan.

Find a practical outcome measure, evidence requirements, and a related calculator for each of the 20 untested Mutual AI Resilience hypotheses.

Design a civilian tabletop exercise or pilot that tests whether AI-assisted cooperation adds value beyond a structured human-led process.

Define the units, denominators and evidence needed to measure service continuity, warning accuracy, human workload and independent recovery.

Plan a two-factor civilian exercise, calculate an interaction from four comparable conditions, and identify when shared dependencies make the result uncertain.

Work through fictional trade and finished-diesel continuity results. Compare three cooperation arrangements and see what changes when a confirmed receipt is late.

Work through fictional backup power, recovery targets, duplicate report tracing and a paired human and AI review exercise with transparent limits.

Explore eight documented trade and fuel cases involving the United States, China, Russia and other participants. See what civilian outcomes a fair test could estimate.

Examine the 2025 U.S.–China tariff pause and a fictional test of how quickly civilian operators can verify a disputed trade notice.

Use China’s 2024 crude oil imports from Russia to examine supplier concentration and the records needed to estimate finished-diesel coverage.

Read the historical record of Russia’s 2023 diesel export restriction and design a civilian exercise for verifying notices and estimating supply gaps.

Review U.S. crude oil exports to China and work through a fictional finished-diesel example with explicit stock, demand, and dispatch limits.

Understand the U.S. distillate days-of-supply measure and learn what local data can estimate a depot’s first missed diesel delivery.

Explore what the Black Sea Grain Initiative documented and how a fair civilian exercise could measure corrections and confirmed cargo receipts.

Follow the 2022 IEA emergency oil releases from announcement to delivery and see what records can confirm usable civilian fuel receipts.

Examine EIA’s 2024 Red Sea petroleum flow data and the local records needed to estimate delays, fuel stocks, and civilian diesel deliveries.

Share, translate, and adapt Safe Nations research, data, software, and graphics for noncommercial purposes under CC BY-NC-SA 4.0.

A complete directory of Safe Nations pages.

Inspect 20 MAR hypotheses with original wording, plain-language explanations, policy questions, safeguards, measurement limits, and proposed tests.

Examine MAR and preventive diplomacy: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and peace agreements: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and positive peace: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and economic interdependence: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and arms control: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and humanitarian cooperation: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and sovereign autonomy: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and shared cyber defense: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and energy continuity: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and food security: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and water cooperation: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and public health: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and climate adaptation: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and disaster preparedness: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and information integrity: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and digital sovereignty: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and scientific diplomacy: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and supply chain resilience: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and anti-corruption & rule of law: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Examine MAR and strategic deterrence: original source text, policy questions, key risk, proposed safeguards, observable test, and evidence limits.

Safe Nations develops an independent Mutual AI Resilience proposal, with transparent research, civilian priorities, policy context, and explicit evidence limits.

Contact Jesse Shelley about practical cooperation to lessen the chance of war between nations. Email jesse@benandanti.org, connect on LinkedIn, and explore Synthetic Imprinting.

Jesse Shelley’s short essay argues that resilient, connected societies make peace a source of power, with cooperation and reducing the risk of war as Safe Nations’ primary goal.

Fourteen sourced cases, led by cooperation between rivals. Explore benefits for both sides and how MAR could reduce miscalculation while preserving national authority.

The Black Sea Grain Initiative illustrates limited cooperation during war: joint verification, civilian food trade, political consent, and a proposed MAR information-support role.

The 1963 Washington–Moscow hotline agreement offers a lesson in reducing miscalculation. Explore a civilian MAR application, benefits to both sides, and limits.

India–Pakistan Indus water cooperation offers lessons in information exchange and dispute management, with the 2025 breakdown explicit and MAR benefits for both sides.

China–India flood-data agreements show a narrow humanitarian channel between rivals. Explore benefits to each side, recorded interruptions, and a proposed MAR exercise.

The Israel–Jordan peace treaty’s water provisions illustrate cooperation between former adversaries, with a proposed MAR role in shared records and dispute clarification.

U.S.–Soviet support for smallpox eradication shows rival nations working on a shared civilian threat. Explore mutual benefits and a bounded MAR coordination analogy.

Apollo–Soyuz shows Cold War rivals agreeing on a limited shared mission. Explore how MAR could support civilian interoperability, mutual benefits, and dialogue.

The IEA’s 2022 emergency oil releases provide a real cooperation example and a starting point for testing MAR support for product matching, delivery timing, and human oversight.

Ukraine–Moldova’s 2022 electricity synchronisation illustrates cross-border resilience, technical safeguards, operator authority, and the limits of AI-supported coordination.

The EU–Ukraine Solidarity Lanes illustrate cross-border freight cooperation, grain-export bottlenecks, customs coordination, and a testable MAR logistics-support role.

A 2024 multinational critical-infrastructure advisory illustrates trusted warning exchange, source provenance, local relevance, and a bounded MAR analyst-support exercise.

International assistance to Türkiye after the February 2023 earthquake illustrates request-led cooperation, offer matching, local authority, and a proposed MAR coordination role.

The 2023–2024 Red Sea shipping disruption illustrates potential MAR cooperation on verified schedules, alternative cargo plans, essential supplies, and local decisions.

The 2021 Colonial Pipeline shutdown illustrates how MAR-style coordination could support verified fuel status, alternative deliveries, and local diesel continuity decisions.

Explore Safe Nations policy briefings for public leaders, a transparent register of political use and support, and separately sourced discussions of related policy ideas.

A civilian cooperation agenda for public leaders: protect essential services, preserve sovereign authority, fund a bounded exercise, and evaluate public value.

A policy briefing on AI resilience for civilian infrastructure, essential fuel logistics, trusted incident reports, and decisions that remain with accountable operators.

How a limited civilian AI cooperation proposal could preserve sovereign decisions, support crisis clarification, and keep the scope of an agreement reviewable.

A practical briefing on public AI accountability: evaluate outcomes, protect civil liberties, account for costs, and establish reasons to stop before expansion.

Generate detailed MAR scenarios for the U.S., China, and Russia. Examine assumptions, compare four briefs, test quantities, and export your work.

Detailed present-day MAR scenarios show what the U.S., China, and Russia could each gain in civilian trade, oil, and diesel, with escalation pathways and limits.

What United States could gain from Mutual AI Resilience in civilian trade, crude oil, and diesel: contributions, retained authority, costs, and escalation limits.

What China could gain from Mutual AI Resilience in civilian trade, crude oil, and diesel: contributions, retained authority, costs, and escalation limits.

What Russia could gain from Mutual AI Resilience in civilian trade, crude oil, and diesel: contributions, retained authority, costs, and escalation limits.

Three governments face conflicting public accounts of civilian diesel availability. They could clarify the facts without creating a three-way fuel deal.

A permitted civilian shipment has inconsistent document versions. Both parties want a reliable handoff and retain separate customs decisions.

A buyer and a permitted supplier have different accounts of an oil commitment. The proposed benefit is a clearer accepted record.

Civilian fuel reports are disputed during a tense relationship. An information exercise need not presume a bilateral import or export.

Several reports repeat one unconfirmed oil claim. Independent source checking could give all three countries a more reliable basis for discussion.

A possible civilian fuel offer does not match an approved receiving requirement. Both sides need qualified review before acceptance.

Participants in permitted civilian activity need to distinguish an accepted diesel commitment from an offer whose conditions remain unresolved.

A severe-weather delay leaves an outdated civilian schedule circulating. A jointly acknowledged update could reduce avoidable commercial friction.

Different definitions of pending, accepted, and delivered make the same civilian commitment appear different across three briefings.

A reporting service fails while officials disagree about the cause. Continuity and attribution need separate treatment.

A civilian food commitment faces an administrative interruption. A limited discussion could clarify needs and status where participation is permitted.

One country stops sharing during a civilian diesel disruption. The process must preserve local capability and show the resulting uncertainty.

Run 26 civilian resilience and policy calculators. Compare stated assumptions, inspect worked examples and explore the 96-module registry.

Find 26 calculators for service continuity, backup power, delivery capacity, forecast accuracy, costs and policy review. Each includes a worked example.

Replicate eight supplied fictional demonstration packs for inventory, policy cost, evidence lineage, fallback, coverage, consent, training cost and interruptions.

Conserves a single physical stock over at most 365 declared daily periods. Offers, accepted commitments and physically received quantities remain separate. See a worked example, input units and limits.

Finds the smallest capacity in a required serial chain and all tied limiting stages. See a worked example, input units and limits.

Adds distinct baseline and incremental costs, subtracts separately declared funding offsets and discounts the declared annual ledger. See a worked example, input units and limits.

Reconciles supplied resources, existing commitments and proposed spending within separately declared accounts for one period. See a worked example, input units and limits.

Trace declared report dependencies to original records without treating copied reports as independent verification. See a worked example, input units and limits.

Keep observation dates, retrieval dates, revisions and forecasts distinct under a declared assessment date. See a worked example, input units and limits.

Evaluates a supplied, dimension-checked linear expression at low, base and high input settings; calculates its exact independent-box range. See a worked example, input units and limits.

Identifies alternatives dominated on every declared comparable dimension without adding unlike units or choosing a policy winner. See a worked example, input units and limits.

Reports disclosed aggregate group coverage and differences, suppresses small cells and prevents overlapping group totals. See a worked example, input units and limits.

Propagate declared availability through a public or fictional function graph and independently evidenced substitutes. See a worked example, input units and limits.

Compare measured manual work with AI drafting, human review and correction for the same explicitly identified task. See a worked example, input units and limits.

Describe dated alerts, audited misses, process latencies and explicitly disjoint service interruption counts. See a worked example, input units and limits.

Compute provider and declared common-parent shares, fractional HHI, and a supplied provider-removal exercise. See a worked example, input units and limits.

Checks explicit authority declarations and required human signoffs for declared activity types. See a worked example, input units and limits.

Reviews structured policy clauses, declared conflicts and scope matches, validates a prerequisite DAG, and calculates conditional dates. See a worked example, input units and limits.

Lists research-design documentation gaps and calculates explicitly comparable observed rates. See a worked example, input units and limits.

Preserves the original suggestion, subsequent human decisions, supplied input references, and visible audit gaps. See a worked example, input units and limits.

Accounts for declared participant resources, local minimums, needs, and distinct authorized, accepted, allocated, and received assistance. See a worked example, input units and limits.

Track a service queue as new requests arrive and declared processing capacity changes. See a worked example, input units and limits.

Account for battery energy, reserve, inverter power, auxiliary draw and load over a declared outage schedule. See a worked example, input units and limits.

Compare verified restoration and the latest consistent recovered record with declared recovery targets. See a worked example, input units and limits.

Calculate the throughput ceiling across directed parallel delivery routes with shared downstream capacity. See a worked example, input units and limits.

Compare two already issued forecasts on the same observed targets using absolute error, squared error and signed bias. See a worked example, input units and limits.

Summarize demand-weighted coverage and hours below a declared minimum across a service interruption. See a worked example, input units and limits.

Compare stated setup and per-case costs over the same workload and find the first whole-case cost crossing. See a worked example, input units and limits.

Calculate the smallest opening buffer that can meet all declared demand under a schedule of early and late confirmed receipts. See a worked example, input units and limits.

Search 96 policy method cards across 16 categories, with input requirements, working core links, planned specializations and explicit validation limits.

Campaign policy researchers can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Candidates can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Legislators can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Presidents and executive advisers can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Diplomats can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Economists can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Energy and trade policymakers can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Agricultural policymakers can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Health officials can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Cybersecurity and AI policymakers can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Environment, climate and water can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Education and labor policy can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Justice and public accountability can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Local governments can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Humanitarian and disaster planning can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Citizens, journalists and researchers can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Cabinet agencies and public administrators can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Academic and independent researchers can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Citizens and civic watchdogs can inspect six relevant policy modules, trace public claims, compare resources and disclose assumptions using transparent, nonpartisan methods.

Compare up to four local policy calculations and three equal-input approaches, with native units, material-condition checks and no assumed AI uplift.

Create a printable civilian executive brief covering known facts, conflicting claims, unknowns, authority, options, costs, fallback and human reviewers.

Read the workbench data contract, five evidence classes, provenance rules, computational test boundaries and reproducible validation requirements.

Review authoritative source documentation, public-data adapter limits, provenance fields and an optional validated World Bank annual-data import.

Understand AI resilience, resilient AI systems and AI-assisted continuity. Explore definitions, failure modes and the proposed Mutual AI Resilience framework.

See how AI can support business continuity and adaptation through data, supply chains and recovery, with practical examples and clear limitations.

Explore AI support for supply chain disruptions, inventory and continuity planning, with civilian examples, visible assumptions and human oversight.

Examine AI-assisted incident triage and recovery planning, including backup verification, human approval, recovery objectives and fallback.

Connect responsible AI and governance to continuity through accountable decisions, data controls, change review and tested fallback procedures.

Plan resilient AI services with dependency mapping, tested fallback, provider independence, recovery objectives and controlled changes.

Assess AI resilience through repeatable tests, operational metrics and recovery checks. Distinguish scenario results from live monitoring.