ScenarioSuperintelligence control and containment assurance

Superintelligent System Containment Researcher

Also called ASI Containment Researcher

A containment researcher would test whether enforceable limits can hold against broadly superhuman AI. The role depends on how AI develops, and it only matters if such systems appear and institutions need credible control.

What they do

The researcher would define what a boundary around a highly capable system actually means, then test whether it can be enforced. They would write precise threat models, state the assumptions each containment claim rests on, and look for ways a capable adversary could break them. They would apply formal methods where a property can be specified, and use scenario analysis where it cannot. They would also record the conditions under which containment fails or cannot be shown. Much of the day would be spent in review with security, control and governance colleagues, challenging each other's arguments.

What a strong hire delivers

In the first months, a strong hire would produce a clear threat model and a written list of assumptions that every containment claim depends on. They would deliver a first set of boundary properties checked under those assumptions, and a documented set of failure and impossibility conditions. They would also give leadership a plain account of what can and cannot be claimed to outside overseers.

How you can tell they are good

Ask them to take one containment claim and set out every assumption behind it. A strong candidate will name where the claim breaks without being prompted. Ask what they would accept as evidence, and what they would refuse to sign off. Good candidates separate what is proven, what is argued and what is hoped. Be wary of anyone who promises certainty.

Why it is hard to fill

If this scenario arrives, the people would be very scarce. No established career path exists, and no title reliably signals the skill. The work mixes formal verification, security, control research and system safety, and few people have worked across all four. Many strong candidates would sit in one discipline and would need to show they can reason about adversaries more capable than themselves.

Where to find them

Today they would come from AI control research, security research, formal verification, and safety engineering for complex systems. Look in frontier lab safety teams, independent safety organisations, security research groups, and academic verification labs. People in AI governance and assurance, or in scenario analysis, could supply the framing and evidence standards, though they would need technical partners.

Interview questions for a Superintelligent System Containment Researcher

  1. Describe a security or safety boundary you tried to verify. What did you assume, and what broke those assumptions?

    A strong answer names real assumptions and a real weakness found. It shows how they scoped the claim honestly. Worry if they describe only successes or treat the boundary as simply secure.

  2. How would you define containment for a system far more capable than its overseers?

    Good answers separate the boundary, the adversary model and the observable properties. They admit limits on what can be defined. Worry about vague language, or definitions that quietly assume the system is cooperative.

4 more questions, with what a strong answer covers, are in the Role Kit.

Job description you can post

We are building research capacity to assess whether enforceable limits can hold around highly capable AI systems. You would work with control, security and governance colleagues to produce containment research that is precise about its assumptions and honest about where it fails.

What you will do

  • Define boundary properties and threat models for highly capable systems
  • State the assumptions behind each containment claim
  • Apply formal methods where properties can be specified
  • Test claims against capable adversary scenarios
  • Document failure and impossibility conditions
  • Brief leadership and external overseers on what can and cannot be claimed

What you will bring

  • Demonstrated work in AI control, security, formal verification or system safety
  • Experience writing threat models or adversarial analyses
  • Evidence of reporting negative results clearly
  • Ability to separate proof, argument and speculation
  • Clear writing for technical and non-technical readers

Nice to have

  • Experience with assurance or oversight processes
  • Background in scenario analysis
  • Published or reviewable research

Free to use and adapt. If it saves you time, a link back to this page is appreciated.

Depends on a scenario

Superintelligent systems with major strategic capability

What AI would need to do
Systems substantially exceed human expert groups across broad consequential cognitive tasks.
What organisations would need to decide
Such systems are developed or deployed and institutions seek credible control and external oversight.
What would rule it out
No broad superhuman capability or no usable, governable deployment emerges.

We give no date. The role becomes real if both conditions hold.

Skills that matter

  • AI controlAdvancedLimit harm from potentially untrusted AI behaviour
  • Formal verificationAdvancedProve stated properties under explicit assumptions
  • AI governance and assuranceWorkingAssign accountability and verify lifecycle controls
  • Frontier scenario analysisWorkingExplore conditional outcomes without claiming predictions

Other roles in superintelligence control and containment assurance

ScenarioSuperintelligence Control Assurance ArchitectA Superintelligence Control Assurance Architect would judge whether claims of control over broadly superhuman systems hold up. It depends on how AI…ScenarioSuperintelligence Intervention Protocol DesignerA Superintelligence Intervention Protocol Designer would specify who may intervene in a broadly superhuman system, how, and whether it would work…ScenarioSuperintelligence Strategic Risk AnalystA Superintelligence Strategic Risk Analyst would map how broadly superhuman systems could affect institutions and security. Whether the job exists…

Related roles in other families

ScenarioGeneral-Purpose Autonomy Assurance ArchitectGeneral-purpose autonomy assuranceA General-Purpose Autonomy Assurance Architect would design how an organisation gains justified confidence in systems given broad, consequential…ScenarioCross-Domain Capability Certification DesignerGeneral-purpose autonomy assuranceA Cross-Domain Capability Certification Designer would define what a general-purpose AI system can credibly be said to do across consequential…ScenarioAutonomous Delegation Accountability OfficerGeneral-purpose autonomy assuranceAn Autonomous Delegation Accountability Officer would keep track of who authorised what when AI systems carry out broad work. The role depends on…ScenarioGeneral-Purpose System Decommissioning PlannerGeneral-purpose autonomy assuranceA General-Purpose System Decommissioning Planner would prepare the controlled withdrawal of autonomous systems that run deeply inside an…

How we wrote this

Role definitions are our own editorial work, checked against occupational classifications, standards, research and employer pages. Titles vary by company, and this is a description of the work, not a job advert. Scenario roles describe responsibilities that depend on stated conditions; they are not forecasts.

← Superintelligence control and containment assuranceAll new job families

Start a search

Hiring a Superintelligent System Containment Researcher?

Tell us the work. We find the people already doing it, whatever their job title says, and you meet them already interviewed for your role.