Where AI governance meets operational reality.

RESEARCH

Research the claim. Examine the evidence.

ODA3 Institute examines AI security questions through public-source analysis, research methods and explicit evidence limits. The aim is to establish what the available record supports—and what remains unresolved.

RESEARCH PURPOSE

Start where the evidence leaves a consequential question open.

Our research agenda connects security requirements, incident records and system behaviour. These questions guide enquiry; they are not claims of completed studies or validated deployment outcomes.

01 · Controls and claims

Does the record support the claim?

Examine the difference between a mapped requirement, an implemented control and evidence of its operation in a defined context.

02 · Incidents and response

What happened—and what can be established?

Examine incident descriptions, response records and source limits while separating reported events from verified observations.

03 · Physical AI

When does a test result carry beyond the test?

Examine configuration, measurement and physical observations before extending a simulation or controlled test to a deployment claim.

Explore Physical AI evidence →

The assurance model explains how requirements, controls and evidence connect to a bounded conclusion.

HOW WE EXAMINE A RESEARCH QUESTION

From a defined question to a finding the evidence can support.

01 · Define

Set the question and scope.

  • State the research question and intended decision.
  • Identify the system, context and observation period.
  • Define criteria and known limits.

02 · Examine

Trace the evidence and test the claim.

  • Establish sources and provenance.
  • Verify claims against defined criteria.
  • Examine conflicting and missing evidence.

03 · Report

Keep the finding within its evidence.

  • Present findings and their stated basis.
  • Record limits and unresolved questions.
  • Document Notably Absent observations.

This describes the research workflow. The assurance model retains its Define → Substantiate → Conclude phases.

Explore the assurance model →

FROM RESEARCH TO SPECIFICATION

Turn research findings into examinable methods.

Research findings can inform methods, evidence schemas and specifications where the evidence supports a repeatable approach. Each published output should state its scope, review status and limitations. Publication alone does not establish operational effectiveness.

Explore Frameworks →

EVIDENCE SOURCING

Source tier describes origin, not usefulness.

Each publication should state the source basis appropriate to its claims. A tier does not by itself determine relevance, integrity or sufficiency for a particular assurance question.

SOURCE CLASS

T1 · Primary Verified

Primary material examined and verified against defined criteria.

SOURCE CLASS

T2 · Secondary Verified

A secondary record with attributable source and documented verification state.

SOURCE CLASS

T3 · Controlled Simulation / Academic Proxy

A controlled or proxy environment; generalization to deployment must be bounded.

SOURCE CLASS

T4 · Anecdotal / Unverified

Reported material whose underlying claim has not been verified against defined criteria.

Where the source tier cannot be established, default to T4 until the underlying basis is documented.

Read the evidence discipline →

FINDINGS

Separate mapping state from assurance conclusion.

Supported, Partial, Gap and Unknown can describe an examined relationship under stated criteria. A report-level result may also be Inconclusive. None of these labels alone certifies a system or proves compliance.

Supported

The stated relationship is sufficiently evidenced for its scope.

Partial

Some elements are supported; material parts remain outside the evidence.

Gap

The necessary relationship or evidence is absent under the criterion.

Unknown

The available record cannot responsibly determine the state.

PUBLISHED RESEARCH

Read the question, sources and limits together.

These selected Insights illustrate questions in our research agenda. Each should be read with its stated source basis and limitations; publication does not itself establish a system-level assurance conclusion.

28 Oct 2026

When Does Simulation Evidence Support Physical AI Deployment?

Examine the conditions and limits that connect a simulation result to a deployment claim.

Read publication ↗
29 Oct 2026

Robot Cancellation Needs Evidence of Physical Motion

A cancellation acknowledgement and the resulting physical motion need separate, traceable evidence.

Read publication ↗
26 Oct 2026

Mapped Is Not Tested: What Boards Should Require From AI Security Controls

A standards crosswalk identifies relationships; operational evidence is needed to examine whether controls work.

Read publication ↗
Browse all Insights →

RESEARCH ENQUIRIES

Bring a bounded research question.

Describe the system or method, the question that matters, available records and known limitations. Explain the intended decision and your authority to provide the material. Please keep confidential records out of the initial enquiry.

Introductory discussions establish fit. Commissioned analysis, study design, validation and advisory work require an agreed paid scope, independence and data-handling checks, written terms and payment before work begins.

For a commissioned threat, incident or sector intelligence brief, explore the paid service scope.Threat Intelligence & Research →

NOTABLY ABSENT

A study does not become assurance by description.

Publication status, source tier, verification, finding state and assurance conclusion remain separate. A controlled reproduction does not establish behavior in an unexamined deployment; an absent record is not evidence of a non-event.