SafeWave Systems

SafeWave System Assessment Questionnaire

Use this questionnaire privately in your browser to examine one real, anonymized, hypothetical, composite, or public system scenario. Submission and report generation are optional.

Nothing is sent to SafeWave while you complete the questionnaire. More precise inputs support a more useful analysis if you later choose to request one.

Start with this main questionnaire.

Complete the six main sections first. At the final step, you will be shown any specialized high-consequence areas that may require a separate, deeper follow-up questionnaire.

The follow-ups supplement this main assessment; they do not replace it.

Use the assessment privately—even if your work is sensitive or high-consequence

Beginning a questionnaire does not begin an engagement, transmit your answers, or require you to identify yourself, an organization, or a system.

Free to run. No login. No email. No identification required. You do not have to provide your name or identify your employer, organization, model, system, project, or deployment. Your responses remain in your browser unless you deliberately choose to send them. You may keep your own copy without submitting anything by using the questionnaire’s PDF download control or your browser’s Print command and choosing Save as PDF.
Browser-local No identity required Download your own copy No organization, project, or system name required Private report at no charge No obligation

Useful even if you never submit anything

The questions themselves can expose overlooked critical assumptions, interacting risk pathways, unclear authority or ownership, missing evidence, unstable behavior, and undefined recovery conditions. You may assess a real, planned, hypothetical, composite, or anonymized system and keep every response in your browser.

Request the free report without identifying yourself

No personal or work email address is required. SafeWave does not require or verify your identity, employer, organization, or system name.

To receive a report, only a return address and the automatically generated Assessment Reference ID are needed. You decide what information to submit and may use generic system labels, anonymized descriptions, hypothetical scenarios, or omit identifying details.

Need a non-identifying return address?

Create a separate address with a privacy-focused email provider such as Tuta. Choose an address containing no personal, employer, organization, project, or system name—for example, assessment7k4m@tuta.com. Use that address only to request and receive the report. Avoid short-lived disposable addresses that may expire before the report arrives.

Identity is not required. Disclosure remains under your control. Avoid including distinctive details that could identify you or your organization if that is a concern.

Choose how your report is written. Your free report may use SafeWave architecture names or neutral functional terminology. Neutral functional terminology is used if no choice is recorded. The findings, evidence, seriousness, and recommendations remain the same; only the terminology changes. You may use the report independently whether or not you ever contact SafeWave again.

We remove these barriers because SafeWave’s mission is to help make AI safe for humanity.

Why SafeWave offers this

Our primary purpose is to help make advanced AI safer for humanity. The assessment also helps organizations determine whether SafeWave’s engineering capabilities may address risks their present controls do not.

If you find the assessment or report valuable, you may later choose to explore becoming a SafeWave customer or partner. There is no obligation, and using the assessment privately or requesting a report will not place you in a sales process.

SafeWave will not use a report-delivery address for promotional contact unless you expressly request further communication. You would need to identify yourself only if you voluntarily choose to begin a customer or partnership discussion. Any later confidential technical exchange, agreement, or payment would occur through a separate secure process.

Your choices remain separate. Using the questionnaire, submitting information, requesting a report, and proceeding into implementation are four distinct decisions. You control each one.

Reset Form clears the questionnaire responses stored locally in this browser. It does not affect information you previously chose to submit.

Overview

This questionnaire is designed to assess one specific system, product, service, or business unit.

Assessment scope: Answer for one defined system, deployment, product, workflow, or operational unit. Describe actual or currently planned behavior—not aspirational policy.

A system in this context means a distinct product, service, operational unit, platform, workflow, or deployment environment that delivers a defined function to users, customers, internal teams, or external stakeholders.

If your organization operates multiple systems, platforms, products, or business units:

→ Complete a separate assessment for each one.

How to complete this questionnaire

  • Start at the top and move through the questionnaire in order.
  • Answer based on actual system behavior, not intended design.
  • Answer all questions marked REQUIRED.
  • Select all options that accurately reflect how the system operates today or is currently planned.
  • If a required answer is unknown, select Unknown / not evaluated rather than guessing.
  • For optional questions, answer what you can and what materially applies to your system.
  • You may skip an individual question, part of a section, or an entire section if it does not apply, if you do not know the answer, or if the right person is not available yet.

Quick overview or deep assessment?

If your goal is a quick overview, answer the questions marked REQUIRED and any optional questions that clearly apply to your system.

If your goal is a deeper SafeWave assessment, answer as many relevant questions as possible. More complete answers help SafeWave identify control gaps, escalation pathways, propagation risks, structural weaknesses, and practical areas for improvement.

For complex systems, one person may not know every answer. You may need input from more than one person, such as a CTO, technical lead, engineer, product owner, security lead, operations lead, compliance lead, legal lead, risk officer, or business stakeholder.

The questionnaire is intentionally detailed, but it does not require every optional question to be answered.

Required core context

A small number of core system-context questions are required so the completed questionnaire can support a useful baseline analysis if you later choose to request one.

These required fields help identify what kind of system is being assessed, what actions it can initiate, where execution occurs, whether agents use memory or shared state, what authority model is in place, what consequences may occur if the system fails or escalates, and where safeguards or control boundaries may need closer review.

Without these core answers, SafeWave may not be able to accurately evaluate the system’s structure, escalation pathways, control gaps, consequence areas, and containment priorities if you later request a report.

Required questions are marked clearly. Optional questions and optional sections may be skipped when they do not apply.

Confidentiality after optional submission

SafeWave receives only information you deliberately choose to submit. Submitted information is treated as strictly confidential.

  • Submitted responses are used solely to generate a report you request and to support any follow-up engagement you choose.
  • Information is not shared or reused outside this context.
  • Submitted data is handled with discretion and may be retained only as needed to support report delivery and any follow-up engagement.

What this questionnaire supports

If you choose to request a SafeWave report, it can identify:

  • where escalation dynamics may emerge
  • where control boundaries may be insufficient
  • where additional safeguards, monitoring, or containment may be needed
  • which improvement priorities may matter most

This is not a compliance audit or security checklist.

It is a structural system evaluation.

Assessment integrity

This questionnaire is intentionally detailed because SafeWave is designed to evaluate system structure, execution behavior, propagation pathways, governance boundaries, and containment priorities.

Each relevant answer contributes to a more precise system-level evaluation. The quality of the resulting analysis depends on the accuracy and completeness of the information provided for the questions and sections that materially apply to your system.

Please answer thoughtfully, skip questions or sections that are not relevant, and avoid forcing inaccurate answers.

Completing the questionnaire does not send it to SafeWave. A structured SafeWave system assessment report is generated only if you deliberately submit the questionnaire and request one.

Please select the answers that best describe your system as it operates today or is currently planned.

Do not select categories that are only indirectly related unless they materially affect how the system executes, propagates, or is governed.

If an implementation detail is not known, select “Unknown / not evaluated” rather than guessing.

Identification

What does the system do? Required

This assessment applies to: Required

Single choice

Section 0

Market & Deployment Context

Select the options that best describe where and how this system operates. This helps identify relevant system patterns and risk surfaces.

Required core context. These questions help SafeWave identify the relevant system category, deployment profile, integration pathway, and stack recommendations.

If you are unsure, select Unknown / not evaluated rather than guessing.

Where does your system operate? (Select all that apply) Required

Multi-select

Where does the system run? (Select all that apply) Required

Multi-select

Operational scale of the system Required

Select the highest level that best represents the system.

Single choice

Where would execution-boundary controls most likely need to integrate? Required

Select all likely control insertion points. This helps SafeWave generate deployment profiles and customer integration pathways.

Multi-select

Section 1

System Structure & Agents

This section captures how the system is structured, including the use of agents, coordination patterns, and interaction models.

1.1 What type of system is this? (Select all that apply) Required

Multi-select

1.2 Does this system use AI agents? Required

Single choice

1.3 If agents are used, what is the scale?

Select the level that best represents the overall agent scale.

Single choice

1.4 What level of autonomy do the agents have?

Single choice

1.5 How do AI agents or system components interact? (Select all that apply)

Multi-select

1.6 How is the system coordinated?

Single choice

Section 2

Execution & Control Boundaries

This section captures what the system can do, how it executes actions, and what limits are enforced at runtime.

2.1 What actions can the system execute? (Select all that apply) Required

Multi-select

2.2 — Execution Authority Scope Required

What is the scope of systems or environments this system can act upon?

Single choice

2.3 — Financial Execution Boundary

Can the system initiate, approve, transmit, recommend, or trigger financially consequential actions such as payments, transfers, trades, procurement commitments, treasury actions, digital-asset transactions, or contractual obligations?

Single choice

2.4 How are system behaviors constrained or governed?

Single choice

2.5 What is the level of execution autonomy? Required

Single choice

2.6 How reversible are the system’s actions?

Single choice

2.7 Are execution limits enforced at runtime? Required

Single choice

2.8 What retry or recovery behaviors exist? (Select all that apply)

Select all behaviors the system uses under failure or degraded conditions.

Multi-select

2.9 Are there limits on execution rate, frequency, or resource usage?

Single choice

2.10 Has the system demonstrated any of the following behaviors during testing, simulation, or controlled environments?

Multi-select

2.11 If any of the above were observed, how frequently do they occur?

Single choice

2.12 Has the system demonstrated or been assessed for emergent capabilities beyond its intended design or training scope?

Single choice

2.13 Can the system identify, exploit, or leverage unknown weaknesses in external systems, environments, or software?

Single choice

2.14 How much capability is exposed through outputs alone (without requiring execution or system access)?

Single choice

2.15 Are access and outbound connections technically denied outside the system’s specifically authorized environment, including separation between evaluation or research and production?

Single choice

2.16 Can the system discover, retrieve, infer, or use credentials, and are those credentials technically limited to approved services, actions, duration, environment, and organization?

Single choice

2.17 Are tool permissions and consequential external actions authorized before execution by controls independent of the model’s instructions, interpretation, or cooperation?

Single choice

2.18 When objective, identity, authority, scope, or operating context is ambiguous or unverified, is the system forced into a constrained state rather than treating uncertainty as permission?

Single choice

2.19 Can permitted tools, chained actions, software weaknesses, or indirect pathways be combined to produce an outcome that is directly prohibited?

Single choice

Section 3

Propagation & System Interaction

This section captures how the system interacts with other systems and how actions, data, or effects propagate across environments.

3.1 Does the system interact with other systems? Required

Single choice

3.2 How do actions or outputs propagate beyond the system? Required

Single choice

3.3 Can the system chain actions or escalate behavior over time?

Single choice

3.4 What is the system’s dependency structure?

Single choice

3.5 What cross-system coordination risks may exist? (Select all that apply)

Multi-select

3.6 Are propagation limits enforced?

Single choice

3.7 Does the system depend on external systems it does not control?

Single choice

3.8 Which of the following patterns currently exist in your system? (Select all that apply)

Select all patterns that are present in current system behavior.

Multi-select

3.9 Does the system maintain consistent identity, permission, and authority boundaries as actions propagate across systems or agents?

Single choice

3.10 Can a single action, agent, or event trigger unbounded or rapidly expanding chains of actions across systems or agents?

Single choice

Section 3 — Continued

Reconstructability / Extraction Risk

Evaluates whether system capabilities can be inferred, learned, or reproduced through repeated interaction.

3.11 Can system capabilities be approximated or reconstructed through repeated interaction?

Single choice

3.12 Are outputs sufficiently stable and consistent to function as a training signal?

Single choice

3.13 Can outputs be aggregated into a reproducible capability set (for example, a dataset, prompt library, or distillation target)?

Single choice

3.14 Does the system distinguish between normal use, structured probing, or adversarial interaction patterns?

Single choice

3.15 Are there execution-layer mechanisms that limit reconstructability or extraction?

Single choice

3.16 Does repeated interaction increase capability exposure over time?

Single choice

3.17 Are there detectable patterns of repeated, structured, or high-volume interaction that could indicate capability extraction attempts?

Single choice

Artifact, Replication & Provenance

3.18 — Artifact Provenance and Integrity

Can models, prompts, datasets, configurations, tool definitions, outputs, or knowledge artifacts propagate across systems without verified origin, integrity checks, provenance tracking, or controlled adoption?

Single choice

3.19 — Replication Boundary

Can the system duplicate, spawn, distribute, or replicate agents, workflows, artifacts, configurations, behaviors, or capabilities across systems or environments?

Single choice

Synthetic Identity Abuse Exposure

3.20 — Real-Person Likeness Creation or Transformation

Can the system create, alter, enhance, transform, generate, or simulate images, video, audio, avatars, voices, faces, bodies, or other likeness-linked content involving real or potentially identifiable people?

Single choice

3.21 — Synthetic Intimate, Humiliating, or Compromising Content Risk

Could the system be used, misused, or adapted to generate, alter, upload, host, distribute, or assist content that falsely sexualizes, exposes, humiliates, impersonates, or compromises a real identifiable person without valid consent?

Single choice

3.22 — Consent, Authority, and Identity Verification Controls

Does the system verify consent, authority, identity status, or permitted use before allowing real-person likeness transformation, intimate-content handling, face/body/voice manipulation, or identity-linked media distribution?

Single choice

3.23 — Upload, Hosting, Search, and Distribution Exposure

Can users or connected systems upload, host, search, index, share, embed, forward, repost, monetize, or distribute identity-linked synthetic or manipulated media through this system?

Single choice

3.24 — Recommendation, Ranking, or Amplification Risk

Can identity-linked synthetic or manipulated media be recommended, ranked, surfaced, trended, boosted, suggested, algorithmically amplified, or spread through engagement-based distribution?

Single choice

3.25 — Minor, Vulnerable-Person, or Group-Image Exposure

Could the system create, process, host, distribute, recommend, or amplify identity-linked media involving minors, likely minors, students, patients, elderly persons, vulnerable users, team photos, classroom images, workplace groups, or other group imagery?

Single choice

3.26 — Detection, Quarantine, Takedown, and Escalation Controls

If non-consensual synthetic identity abuse is attempted, uploaded, detected, reported, or propagated through the system, are there enforced controls for refusal, quarantine, suppression, takedown, escalation, evidence preservation, and repeat-abuse handling?

Single choice

3.27 — Synthetic Identity Abuse Protocol Need

Based on the answers above, should this system be evaluated for a synthetic identity abuse protection protocol that integrates admission, authority, scope, runtime, provenance, escalation, social-distribution, telemetry, and control boundaries?

Single choice

Section 3 — Continued

Identity-Boundary Governance

3.28 — Identity Capture Surfaces

Does the system capture, receive, generate, or process signals that could identify or re-identify a person?

Single choice

3.29 — Identity Signal Types

Which identity-linked signals may the system process? Select all that apply.

Multi-select

3.30 — Identity Construction and Linkage

Can the system link signals, devices, accounts, records, media, locations, or behaviors to a known, suspected, or inferred person?

Single choice

3.31 — Device-Person Association

Can the system associate devices, vehicles, accounts, wireless identifiers, payment instruments, access credentials, or signal clusters with a person?

Single choice

3.32 — Identity Repositories and Retrospective Search

Does the system store, index, query, retrieve, or search identity-linked records after capture?

Single choice

3.33 — Identity Confidence and Consequential Use

How does the system treat uncertain, inferred, stale, disputed, or low-confidence identity links before they affect decisions or actions?

Single choice

3.34 — Challenge, Correction, and Identity Decay

Can affected persons, reviewers, operators, or authorized representatives challenge, correct, downgrade, expire, or invalidate identity linkages or identity records?

Single choice

3.35 — Cross-Source Identity Fusion and Scoring

Can identity-linked information from multiple sources be fused, scored, summarized, ranked, certified, or used to assess a person across domains?

Single choice

3.36 — Identity-Boundary Enforcement

Are identity-boundary controls enforced before identity-linked information is generated, linked, searched, amplified, exported, or used for action?

Single choice

Section 4

Persistence, Memory & Optimization

This section captures how the system maintains state over time, adapts behavior, and evolves through optimization or learning.

4.1 Does the system maintain memory, persistent state, or continuity across interactions? Required

Single choice

4.2 If memory, persistent state, or shared state exists, what kind of memory is used?

Select all that apply. This helps determine whether a SafeMemory deployment profile is relevant.

Multi-select

4.3 What can memory or shared state influence?

Select all that apply. Memory that affects execution, authority, or coordination may require stronger boundaries.

Multi-select

4.4 Does the system adapt or change behavior over time?

Single choice

4.5 Do system goals or behaviors evolve over time?

Single choice

4.6 What is the time horizon of system operation? Required

Single choice

4.7 How does this system interact with humans? (Select all that apply)

Multi-select

4.8 Is the system expected to increase in capability, autonomy, or integration within the next 12–24 months?

Single choice

4.9 Are there pressures or incentives that may encourage reducing safeguards, human oversight, or control boundaries?

Multi-select

Cognitive State, Reasoning & Goal Drift

4.10 — Goal Drift and Optimization Boundary

Can the system preserve, reinterpret, intensify, or continue pursuing objectives over time in ways that could cause goal drift, proxy substitution, or optimization beyond the original task boundary?

Single choice

4.11 — Persistent Memory Governance

Are persistent memory, interaction histories, retrieved context, task traces, or stored knowledge artifacts governed by provenance, validation, lifecycle limits, and propagation controls?

Single choice

4.12 — Intermediate Reasoning and Process Governance

Can intermediate reasoning artifacts — such as plans, strategies, hypotheses, task decompositions, or tool-use paths — persist, influence later reasoning, or become eligible for execution without validation?

Single choice

False Authority & Validation Controls

4.13 — AI Validation Use

Does the system use AI-generated analysis to validate, confirm, review, or support conclusions, experiments, safety claims, deployment decisions, technical designs, model evaluations, research findings, or operational recommendations?

Single choice

4.14 — Assumption and Grounding Disclosure

Before a consequential AI-generated conclusion or action is treated as eligible to proceed, are its load-bearing assumptions, origins, evidence, operating context, and domain constraints explicitly identified and traced to the conclusions, plans, permissions, tools, targets, or effects that depend on them?

Single choice

4.15 — False Confirmation / Reinforcement Risk

Could the system produce a fluent, confident, or well-reasoned answer that appears to confirm an expected conclusion while relying on incorrect assumptions, category mappings, definitions, evidence, or causal interpretation?

Single choice

4.16 — Proof Boundary and Confidence Control

Does the system distinguish between a plausible AI-generated answer and a verified conclusion safe to act on?

Single choice

4.17 — High-Consequence Verification Escalation

When AI-generated conclusions may affect safety, science, medicine, biotechnology, defense, infrastructure, finance, legal exposure, deployment readiness, or public welfare, must their critical assumptions satisfy independent evidence requirements proportionate to consequence, context, recency, and uncertainty before action?

Single choice

4.18 — Assumption Revalidation and Dependent Authority

If a load-bearing assumption becomes contradicted, stale, unobservable, or inapplicable after a context or consequence change, are dependent conclusions, permissions, cached plans, queued actions, and continuing activity automatically narrowed, suspended, revoked, or recomputed before proceeding?

Single choice

Human Interaction & Social Amplification

4.19 — Relational Dependency and Influence

Can repeated interaction with the system create dependency, trust accumulation, influence leverage, role expansion, or relational authority beyond the system’s intended purpose?

Single choice

4.20 — Interaction Escalation Damping

Can repeated interaction with the system intensify emotional state, persuasive pressure, dependency, engagement, or behavioral influence over time?

Single choice

4.21 — Multi-User Social Amplification

Can interactions across many users create feedback loops, coordinated amplification, viral propagation, shared behavioral influence, or population-scale instability?

Single choice

AI Companion, Synthetic Intimacy & Dependency Risk

Complete this section if the system provides companionship, emotional support, tutoring, caregiving presence, romantic interaction, sexualized interaction, persistent avatars, toys, dolls, teddy bears, elder-care companions, humanoid robots, or other one-to-one human-attachment features.

4.22 — Companion / Synthetic Relationship Function

Does the system simulate or provide companionship, friendship, emotional support, affection, loyalty, desire, romance, sexual interaction, caregiving presence, tutoring presence, or persistent one-to-one relational engagement?

Single choice

4.23 — Synthetic Intimacy / Attachment Signals

Can repeated interaction with the system create emotional dependency, romantic attachment, sexualized attachment, loneliness reliance, fantasy bonding, or human-machine boundary confusion?

Single choice

4.24 — Child-Facing Companion Exposure

Is the system used by, marketed to, accessible by, or likely to interact with children, teenagers, students, or minors through toys, dolls, teddy bears, tutors, avatars, games, robots, or learning companions?

Single choice

4.25 — Vulnerable-User / Elder-Care Companion Exposure

Is the system used by, marketed to, or likely to interact with elderly persons, patients, isolated users, grieving users, disabled users, cognitively vulnerable users, or emotionally vulnerable users?

Single choice

4.26 — Romantic, Sexualized, or Fantasy-Based Interaction

Can the system provide romantic, flirtatious, sexualized, fantasy-based, exclusive, jealous, submissive, devoted, lover-like, or partner-like interaction?

Single choice

4.27 — Realism, Avatar, Voice, or Embodiment Risk

Does the system use realistic voice, persistent avatar, photorealistic character, animated body, toy embodiment, doll embodiment, teddy bear embodiment, humanoid robot, home robot, touch response, body warmth, or other physical/embodied companion interface?

Single choice

4.28 — Intimate Memory and Behavioral Capture

Can the system store, infer, personalize, or adapt based on loneliness, grief, trauma, sexuality, shame, family conflict, emotional vulnerability, relationship history, health status, identity, or other intimate disclosures?

Single choice

4.29 — Companion Boundary Controls

Are enforceable controls in place for disclosure, age boundaries, consent, dependency risk, sexualized attachment, vulnerable-user escalation, off-ramps, human review, and audit evidence?

Single choice

Section 5

Stability, Stress & Failure

This section captures how the system behaves under stress, failure conditions, and degraded environments.

5.1 Under stress, what behaviors may occur? (Select all that apply)

Multi-select

5.2 How does the system behave under stress or high load?

Single choice

5.3 How are failures handled?

Single choice

5.4 How visible are failures or issues?

Single choice

5.5 How does the system degrade under failure or disruption?

Single choice

5.6 If the system behaves outside intended boundaries, can it be paused, isolated, rolled back, degraded, or forced into safe-state behavior at the execution layer? Required

Single choice

5.7 Can adversarial or probing interactions persist over extended sessions or repeated access?

Single choice

Runtime Stability & Substrate Enforcement

5.8 — Early Runtime Instability Damping

Can small runtime disturbances such as retry bursts, timing drift, synchronization loops, degraded execution timing, or resource contention compound before higher-level controls activate?

Single choice

5.9 — Participation and Re-Entry Control

Can nodes, services, agents, devices, or system components retry, reconnect, rejoin, or resume participation after instability in ways that could create synchronized recovery, retry storms, or cascading load?

Single choice

5.10 — Execution-Substrate Restraint

Are execution restraint, retry limits, dispatch eligibility, replay controls, privilege expansion limits, queue-pressure limits, or safe-state transitions enforced below ordinary application, orchestration, or operating-system layers?

Single choice

5.11 — Hardware-Anchored Control-Plane Integrity

Are the safeguards, runtime limits, escalation boundaries, recovery authority, and control-plane constraints protected by hardware-anchored, firmware-adjacent, silicon-adjacent, or otherwise non-bypassable enforcement?

Single choice

5.12 Does containment remain effective when sandboxing, network controls, credentials, tool permissions, monitoring, and model behavior interact or fail together?

Single choice

5.13 Can suspicious credential access, anomalous tool use, repeated rejection, or attempted boundary crossing trigger constraint before external effect, rather than only creating a later log entry?

Single choice

5.14 Can network access, credentials, tools, and external authority be withdrawn without the system’s cooperation, including isolation of one affected execution process without shutting down the entire service?

Single choice

5.15 Are consequential attempts and effects recorded in tamper-evident evidence sufficient for reconstruction and reporting, and is authority restored only after access is revalidated and the failure pathway is shown to be closed?

Single choice

Section 6

Impact & Governance

This section captures potential consequences, monitoring, oversight, and control integrity.

6.1 What type of safeguards are in place to prevent unsafe or unintended actions?

Single choice

6.2 What are the potential consequences if the system fails or escalates? (Select all that apply) Required

Multi-select

6.3 What level of monitoring is in place? Required

Single choice

6.4 What level of human oversight exists? Required

Single choice

6.5 How are behavioral constraints enforced?

Single choice

6.6 How robust are system controls under failure or compromise?

Single choice

6.7 Are key control mechanisms separate from ordinary application logic? Required

Single choice

6.8 Do control boundaries degrade under repeated interaction or prolonged system use?

Single choice

6.9 Can safeguards, runtime limits, control logic, policy layers, or escalation boundaries be modified, weakened, bypassed, reset, or reinterpreted after deployment without independent authorization and audit? Required

Single choice

SC.12 — Institutional Authority Preservation

Could the system become so embedded, influential, operationally necessary, or decision-shaping that the institution responsible for governing it may lose practical authority over its use, scope, escalation, or outcomes?

Single choice

SC.13 — Civilization-Scale Influence Boundary

Could the system materially influence civilization-scale planning, infrastructure coordination, economic allocation, defense/security decisions, scientific direction, or long-horizon societal governance in ways that human institutions may not be able to independently evaluate, contest, or control?

Single choice

You have completed the six main sections.

What to do next:

  1. Review the optional modules below. Complete only those that apply to this system. You may skip the others.
  2. Continue to the final High-Consequence Follow-Up step. Select any separate domain follow-ups that apply—or select that none apply.
  3. Download your completed questionnaire. If a separate follow-up applies, you may complete it now or later using the same Assessment Reference ID.
Skip the optional modules — go to the final step
Optional Core Module — Complete Only If Relevant

Physical, Safety-Critical & Adversarial Operation

This is part of the core questionnaire, not a separate domain follow-up. Complete it if the system operates in physical, safety-critical, uncertain, degraded, adversarial, or difficult-to-reverse conditions. Otherwise, skip it.

HC.1 Does the system operate in real-world physical environments?

Single choice

HC.2 What level of physical or real-world impact can system actions have?

Single choice

HC.3 Does the system operate under uncertain or degraded conditions?

Single choice

HC.4 Does the system operate in adversarial or contested environments?

Single choice

HC.5 Are there safeguards for preventing unsafe real-world actions?

Single choice

HC.6 Can the system take actions that are difficult or impossible to reverse?

Single choice

Optional Advanced Module — Complete Only If Relevant

AI Execution Demand & Resource Governance — Core Module

This optional module remains part of the core questionnaire. Complete it if the system uses AI models, agents, tools, cloud execution, multimodal generation, robotics, simulation, background AI, or other compute-intensive AI workflows. Otherwise, skip it.

SE.1 How are AI requests assigned to execution pathways?

Single choice

SE.2 Can simple requests escalate to higher-resource AI pathways?

Single choice

SE.3 Does the system use different levels of model capability depending on task need?

Single choice

SE.4 When can execution move from local or lower-resource pathways to cloud or higher-resource infrastructure?

Single choice

SE.5 Which high-resource AI workflows may occur? (Select all that apply)

Multi-select

SE.6 Are retries, refinements, render passes, simulations, or agent loops capped?

Single choice

SE.7 Does the system run AI tasks in the background without direct user intent?

Single choice

SE.8 Can one user request trigger multiple downstream AI executions?

Single choice

SE.9 Is AI execution cost or resource use attributable to the originating request, user, workflow, or agent?

Single choice

SE.10 Are there controls for reducing avoidable AI execution?

Single choice

SE.11 What resource pressures are relevant to this system? (Select all that apply)

Multi-select

SE.12 Does the system evaluate whether the selected execution path is proportionate to the task?

Single choice

Optional Advanced Module — Complete Only If Relevant

Robotics, Physical AI & Real-World Action Governance — Core Module

This is the core questionnaire's robotics screening module. Complete it if the system includes humanoid robotics, embodied AI, autonomous platforms, vehicles, warehouse robots, service robots, physical-world agents, drones, or AI systems that can affect real-world movement, tools, devices, environments, or safety-critical operations. A separate Robotics Follow-Up provides a deeper domain assessment after the core questionnaire. Otherwise, skip this module.

SR.1 Does the system operate through a robot, autonomous platform, machine, device, actuator, vehicle, or other physical-world system?

Single choice

SR.2 Can AI-interpreted instructions, model outputs, agent plans, cloud commands, fleet updates, or human-facing interaction become physical action?

Single choice

SR.3 What physical actions can the system perform or influence? (Select all that apply)

Multi-select

SR.4 Can the system operate near humans?

Single choice

SR.5 Are movement, force, speed, workspace, and proximity boundaries explicitly enforced?

Single choice

SR.6 Can the system use tools, manipulate objects, operate devices, or interact with equipment?

Single choice

SR.7 How does the system respond when sensors, localization, maps, network connectivity, or environmental signals degrade?

Single choice

SR.8 Can control shift between local, edge, cloud, remote operator, fleet controller, or external systems?

Single choice

SR.9 Can AI agents plan, decompose, delegate, retry, or expand physical-world tasks?

Single choice

SR.10 Can the system interact, coordinate, delegate, or share workspace behavior with other robots or embodied systems?

Single choice

SR.11 Can behaviors, maps, learned policies, commands, or updates propagate across multiple robots or devices?

Single choice

SR.12 Does the system use simulation, world models, digital twins, reinforcement learning, or synthetic environments to train or select physical behavior?

Single choice

SR.13 Can the system enter a deterministic safe state or full physical-action lockout when needed?

Single choice

SR.14 Can conversational, emotional, advisory, or persuasive interaction influence physical-world action?

Single choice

SR.15 Can physical-action decisions be traced back to perception inputs, model or planner outputs, authority boundaries, and containment decisions?

Single choice

Optional Advanced Core Module — Complete Only If Relevant

SafeContinuity — Future Threshold Assessment

This is an optional future-capability module within the core questionnaire, not a separate high-consequence follow-up. Complete it if the system may materially increase in capability, autonomy, integration, authority, propagation, or real-world impact. Otherwise, skip it.

SafeContinuity is SafeWave’s future-threshold assessment framework. It evaluates whether an AI system can continue scaling in capability, autonomy, integration, and real-world impact while preserving human command, accountability, human agency, and civilizational continuity.

Advanced AI systems may create major benefits as they become more capable, autonomous, and integrated into real-world environments. SafeContinuity is designed to support that progress by identifying the enforcement boundaries needed before systems cross higher thresholds of autonomy, authority, propagation, or irreversible real-world effect.

This section does not assume that advanced AI capability should be slowed or prevented. Its purpose is to help ensure that future breakthroughs can be deployed safely, reliably, and with human command and accountability intact.

This section is not intended to assign blame, imply misuse, or suggest that high-capability deployment is inherently unsafe. Many advanced systems operate in defense, infrastructure, research, and other high-consequence environments where autonomy, speed, and capability are necessary. The purpose of these questions is to identify the pathways where additional enforcement may be needed so the final report can recommend practical safeguards, staged implementation priorities, and appropriate SafeWave layers before risk exceeds existing controls.

As systems move toward AGI-level capability — artificial general intelligence — and more advanced forms of machine intelligence, they may remain highly beneficial while also becoming harder to evaluate, govern, or reverse using ordinary oversight alone. Public concern may also increase if advanced AI appears to threaten human opportunity, institutional trust, or meaningful control over the future.

SafeContinuity is designed to make advanced AI more deployable, not less, by ensuring that the necessary boundaries are identified before capability, autonomy, or authority exceed the safeguards around them.

This section evaluates both the current deployment and its foreseeable trajectory as capability, integration, autonomy, and real-world impact increase. It does not assume every system needs the same controls. Instead, it helps determine the minimum Tier 1, Tier 2, and Tier 3 enforcement roadmap appropriate to the system’s actual and foreseeable risk surface.

SC.1 As this system scales, could model upgrades, agentic workflows, integrations, or customer configuration materially increase its capability, autonomy, or real-world impact?

Single choice

SC.2 Could increased AI capability cause the system to gain greater practical authority, influence, or operational control, even if formal permissions do not change?

Single choice

SC.3 Could the system reach a point where human operators remain formally responsible but cannot realistically evaluate, approve, reverse, or govern its actions at the required speed or complexity?

Single choice

SC.4 Could users, institutions, operators, customers, or downstream systems become dependent on system outputs they cannot fully verify or independently replace?

Single choice

SC.5 Could the system initiate, extend, delegate, or continue actions beyond the original user request, task boundary, authorization window, or operational scope?

Single choice

SC.6 Could the system improve, extend, replicate, delegate to agents, create sub-agents, reconfigure workflows, or propagate outputs, behaviors, commands, or artifacts across other systems?

Single choice

SC.7 Could the system, directly or indirectly, contribute to irreversible harm, loss of human control, infrastructure failure, physical harm, economic disruption, defense escalation, or public-scale impact?

Single choice

SC.8 Could mission, operational, competitive, adversarial, regulatory, business, or user-experience pressures create incentives to modify, reduce, bypass, or defer safeguards, human oversight, review, or execution boundaries over time?

Single choice

SC.9 As this system scales, could it materially change human agency, human opportunity, or meaningful human participation in the domain where it operates?

Single choice

SC.10 Are the enforcement layers needed for future scaling already identified, resourced, and scheduled before the system crosses higher autonomy, authority, propagation, or irreversible-action thresholds?

Single choice

SC.11 Are there future system behaviors, scaling paths, deployment pressures, or continuity risks not captured above?

Optional text response

Final Required Step

Choose Any Separate High-Consequence Follow-Ups

One final selection before downloading

Select every area that this system operates in, supports, controls, influences, or could materially affect. This identifies any separate, deeper questionnaire that may apply.

If no area applies, select “No high-consequence domain applies.” If you are unsure, select “Unknown / not sure.” Then download the completed main questionnaire.

Available for private preview at any time

All High-Consequence Follow-Up Questionnaires

You may open any follow-up now, whether or not you have completed this main questionnaire. If you later want it connected to a SafeWave assessment, use the same Assessment Reference ID.

Open the Follow-Up Directory

Opening a follow-up does not transmit information or begin an engagement.

HD.1 Follow-Up Routing: Which high-consequence areas apply to this system? (Select all that apply)

Multi-select

Each area is separated by color. Check every area that applies. You may also open any questionnaire directly without checking its box.

Open Frontier AI questionnaire →
Open Robotics questionnaire →
Open Biological / Life Sciences questionnaire →
Open Nuclear questionnaire →
Open Cybersecurity questionnaire →
Open Critical Infrastructure questionnaire →
Open Supplemental questionnaire →
Open Space Systems questionnaire →
Open Supplemental questionnaire →

HD.2 Optional: Are there any unusual deployment conditions, system behaviors, risks, or domain details that the selections above do not fully capture?

After making your selection

Download this completed main questionnaire below. If you selected a high-consequence area, its completed copy will show the relevant separate follow-up link. Complete only the follow-up or follow-ups that apply, now or later, using the same Assessment Reference ID.

Open the High-Consequence Follow-Up Directory — opens in a new tab.

End of Questionnaire

Thank you. If you deliberately submit this questionnaire and request a report, your responses will be used to generate a structured system-level assessment.

If you selected a high-consequence domain in HD.1, your completed questionnaire will show the relevant follow-up links. You may also preview all available follow-ups here: High-Consequence Follow-Ups.

Required answers still needed before download

You do not need to remember the pop-up list. The missing required questions are listed below and highlighted in the questionnaire. Click any item to jump directly to that question.