AI agents investigate every alert, add context, and prepare cases for response, on a full-stack SecOps platform that captures and correlates all your security data.
Cloud-based SIEM platform providing visibility across all your data sources
Every alert investigated, with a verdict and the evidence behind it
Manage your external footprint, publicly available data, and Dark Web leaks
Log management and threat detection across any of your log sources
Get ultimate assurance that your environment is free from threat actors
Fully featured distrubuted NDR for on-premises and public cloud environments
A SecOps platform purpose-built for industrial and OT environments
Integrated endpoint agent providing deep visibility, response and forensics
Deploy honeypots and honeytokens and stay ahead of threat actors
24x7 security operations from our Australian SOC. AI investigates every alert, and experienced analysts validate findings and contain threats around the clock.
24x7 managed detection and response across endpoint, network, cloud, identity, and log sources
AI investigations and 24x7 MDR on the Microsoft Sentinel you already own
Eyes-on-glass SOC and continuous threat hunting
Expert hunters and AI agents find attacker activity your alerts missed
Quickly recover from cyber attacks and breaches
ThreatDefence is the only SecOps as a Service company providing broad coverage across your entire technology stack with evidence-based security.
We provide security teams with full-stack SecOps infrastructure – you get deep visibility from day one, gain control over your security data, and get access to a comprehensive set of ready-to-use tools, detections, workflows, playbooks and scenarios.
24x7 SOC, Continuous Monitoring, Threat Containment and Incident Response
Visibility and actionable insights for all IT assets
Leverage our Cyber Range to train defenders based on real-world scenarios and simulations
OT and industrial cyber security services for critical infrastructure
Monitoring, evidence and a 24x7 SOC in one platform
Compliance monitoring, evidence and board reporting
Privacy Act reform impacts, obligations, and practical security considerations
Monitoring and compliance support for critical infrastructure environments
A practical guide to securing critical infrastructure with continuous monitoring
A practical buyer’s guide for schools evaluating SIEM and SOC services
A practical guide to selecting the right SIEM and SOC service for local government
Articles on security operations, AI and compliance
Events, awards and company news
Read what our customers say about us
Solution briefs and datasheets
How we keep your data secure
Technical support and knowledge base
Guides, case studies, and advisories in one place
How AI agents investigate alerts and run security operations
Free ebook: a practical guide to adopting an AI SOC platform
A practical guide to incident response and digital forensics
Where SOAR playbooks fit and where AI agents take over
Free ebook: modern security operations with Microsoft Sentinel
How the framework is structured and what changes from the Essential Eight
November 13, 2026
November 10, 2026
November 6, 2026
We provide end-to-end SecOps solutions for Managed Security Service Providers (MSSPs), Managed Service Providers (MSPs), and Value-Added Resellers (VARs).
Find out how you can become a ThreatDefence partner.
Protect your entire technology stack and build your SecOps on your own pace
White-labeled AI SOC for every client tenant, co-managed with our SOC team
Launch your SOC business without any upfront investment
Embed security operations, threat intelligence and exposure data in your product
Sovereign SOC for Australian MSPs and MSSPs, with onshore analysts and client data
Set the rules for closing alerts automatically, with evidence checks and review
How MSPs deliver branded security monitoring with a SOC partner
ThreatDefence was an official partner of the Cyber Resilience Summit NSW, 29 October 2026 at Hilton Sydney.
Our team was back at BSides Canberra, and our cyber quiz was a hit again.
Two days in Perth with our partner Ever Nimble, on security operations and cyber resilience.
Sponsors of the AUSCERT Cyber Security Conference on the Gold Coast.
Sponsors of the Gartner Security & Risk Management Summit in March.
Sponsors of the Cyber Security for Critical Assets Summit in Perth, February 2026.
Expert conversations with speakers and attendees at CyberCon Melbourne 2025.
An autonomous SOC uses AI agents to investigate alerts and perform authorized actions without waiting for approval at every step. People define the boundaries, oversee results, and take responsibility for decisions that affect the organization.
Autonomy means software can make selected decisions and carry out work within delegated authority. An agent might choose its next investigative query, close a case that meets an approved policy, or execute a permitted containment action.
This changes where analysts enter the work. Instead of starting with an unexplained alert, they can review an investigation already supported by evidence, with a response prepared. Routine work can proceed without waiting in the queue, while uncertain or consequential cases reach a person.
The boundaries matter more than labels such as self-driving SOC or lights-out SOC. Establish three permissions separately:
Which data and tools can the agent access, and what may it do to gather evidence?
May it classify the alert, change severity, close the case, or only recommend an outcome?
Which changes may it execute, against which targets, and under what conditions?
An agent can investigate independently while every closure and response action still requires approval. An agentic SOC describes how agents carry out work; autonomy describes how much authority they have to proceed without a new human decision.
Autonomous SOC levels help teams discuss the division of work. The categories below are a planning aid, not a universal industry standard or certification. Vendors use different names and boundaries.
Apply the categories to workflows and action types rather than assigning one level to the whole SOC. The same organization may authorize automatic investigation across its tools, automatic closure for selected benign findings, and human approval for server containment.
The fully autonomous category describes an unattended end state. It is not a prerequisite for getting value from AI. Removing delays from evidence gathering and routine decisions can expand capacity while important business decisions stay with people.
Two common arrangements place the human decision at different points.
Human in the loop means a specified decision or action waits for an authorized person’s approval. The analyst needs the evidence, target, proposed change, expected effect, and recovery options to make that approval meaningful.
Human on the loop means pre-authorized work proceeds under supervision. People review outcomes, investigate exceptions, and can intervene or suspend further execution.
Both require an operating process. An approval queue needs someone available to work it, with escalation if a request is unanswered. Supervision needs clear exception signals and a review schedule. A dashboard nobody checks provides little oversight.
Keep people responsible for high-impact decisions, incident leadership, and external communications. Agents can assemble evidence and prepare recommendations, while the appropriate owner decides on service shutdowns, regulatory notifications, and other material actions.
Assess the action and its target together. A correct finding does not establish that every proposed response is appropriate.
Consider the harm of waiting, the harm of acting incorrectly, the evidence available, and the systems that depend on the target. Also consider whether the task occurs often enough to justify implementation, testing, and ongoing review.
These policies are illustrative starting points, not universal defaults:
The same isolation request can have different consequences on an employee workstation and a server running a critical service. A password reset for a human user may interrupt one person; a change to a service account may affect several applications.
Recovery also needs a precise meaning. Releasing an isolated host can restore connectivity, but cannot recover lost work. A revoked session cannot be restored as the same session. Reopening an incorrectly closed case cannot undo the time an attacker gained.
Use reviewed outcomes to establish what a confidence score means for the specific workflow. A score should never override a protected-target exclusion, an approval requirement, or missing evidence.
Closure criteria also need to match the conclusion. A malicious phishing message remains malicious even if nobody clicked it. Lack of observed harm can affect response, but does not make the original activity benign.
An agent needs enough information to assess both the activity and the consequences of its proposed action. Connect the required sources and check that they return current, usable records. A failed query or missing source must remain visible.
Keep business context current: asset ownership, critical services, account roles, approved administration, maintenance windows, and dependencies. Technical telemetry may show that an account changed a setting without explaining whether the change was authorized.
For example, an agent may correctly identify suspicious activity on a workstation but propose isolation without knowing that the device supports a critical operational function. The missing context changes the response decision even when the threat assessment is sound.
Define what happens when that information is unavailable. The workflow can continue gathering permitted evidence, hold an action, or escalate. It should not treat an absent restriction as permission to proceed.
The AI SOC platform connects agent decisions to real tools and systems. Its execution controls determine whether a proposed query or response is allowed. Instructions telling the model to behave carefully are only one part of that design.
Use separate controls for:
Restrict tools, data, credentials, and tenant scope to the task. The model must not be able to expand its own permissions.
Validate the target, operation, approval, exclusions, and limits before execution.
Cap targets, action rates, concurrent runs, and resource use so one faulty decision cannot expand unchecked.
Preserve queries, relevant results, verdicts, approvals, and action outcomes for review.
Let authorized people reduce permissions and suspend workflows, with pending and in-progress work visible.
Treat email, document, file, and log content as untrusted evidence. An attacker may place instructions in that content to influence a verdict or misuse a tool. Filtering suspicious text helps, but access and approval requirements must still hold if the model follows a misleading instruction.
Reviewed context and memory need similar care. An incorrect case conclusion should not become an unquestioned fact in later investigations.
Tools, tenant scope, approvals, and limits.
Query data and run approved actions.
Tests explanations and picks the next step.
Evidence rules and escalation guidance.
Case and validated tenant context.
Isolation, credentials, network, and resources.
Keep these controls consistent across agents and direct analyst interactions. Calling the same tool through a conversational agent must not create a route around tenant permissions or approval rules.
Execution needs its own checks after the investigation is complete. Revalidate an action that has waited for approval if the target, policy, or relevant circumstances have changed.
After execution, verify the resulting state where the tool supports it. An accepted API request may represent a queued job. A timeout may leave the outcome unknown rather than prove that nothing happened.
Check before retrying. Preserve successful and failed steps separately, and hold dependent work when partial completion makes the next action inappropriate.
Test suspension and recovery alongside successful execution. Suspending a workflow should stop further authorized work from being issued, but a request already sent to another system may still complete. The team needs to inspect that state and decide what recovery is required.
Assign cases returned to manual handling to a person. A stop control protects execution only if unresolved threats continue receiving attention.
Choose frequent work with reliable data, clear boundaries, and outcomes analysts can check. Begin with investigation and recommendations while people retain closure and response decisions.
Compare agent findings with reviewed live cases and historical incidents with known outcomes. Include legitimate activity that looks suspicious, confirmed attacks, conflicting records, and missing sources. Review every material disagreement to establish whether the cause was data, context, investigation logic, or an incorrect reference assessment.
Once the investigation performs reliably, introduce specific actions through approval. Test target selection, execution, notification, and recovery using controlled systems. Accurate recommendations alone do not demonstrate that response integration works.
Authorize a narrow subset to run independently only when the evidence supports it and the designated owner approves. Keep other actions at their existing settings. A successful workflow does not justify increasing authority across the entire SOC.
Track whether autonomy improves operations without concealing errors. Review results by workflow, alert class, environment, and action type:
Evidence completeness, analyst corrections, incorrect closures, and missed escalations found through review.
Verified completion, failed or unknown outcomes, recovery activity, and unintended disruption.
Investigation time, approval waiting, backlogs, and analyst effort saved after review and maintenance.
Whether uncertain cases reach the right person and receive a timely decision.
Analyst agreement is useful, but agreement is not independent proof of correctness. Review supporting evidence and verified outcomes where available. Sample automatic closures, including new or changed classes, and record how the sample was selected.
Use feedback to improve context, skills, and detections. Changes to a model, tool, permission, or investigation method can also alter behavior, so previous results need to be reassessed where those changes matter.
Version autonomy settings and retain the evidence supporting them. Each workflow needs a named owner, approved scope, required sources, escalation conditions, target exclusions, execution limits, and a review date.
Increases require approval from the customer or designated service owner. Permissions can also move down. Repeated data gaps, incorrect closures, unexpected actions, or changes to critical systems may justify returning a workflow to approval or recommendation-only mode.
Preserve the investigation record when a workflow is suspended. Determine the cause, correct it, and validate the affected behavior before restoring execution permissions.
This makes autonomy an operational decision that can be inspected and changed. The useful measure is how much reliable work the SOC completes, with clear ownership of what still requires a person.
ThreatDefence combines an AI SOC platform with a 24x7 SOC. AI triage investigates every alert and records a verdict with confidence, evidence, the queries run, and recommended actions.
Autonomy is configured per customer and action type in three modes:
The agent investigates and a person decides.
The agent prepares the action and holds it for approval.
For nominated action types, the agent contains first and reports afterward.
Explore the AI SOC Platform
Start with one high-volume workflow in recommend-only mode and compare findings with analyst assessments. Expand its authority when reviewed results justify the change, keeping a record of who approved it.
A demo shows operating modes, approvals, and quality reporting on a live tenant.
See operating modes per action type, approvals, and quality reporting on a live tenant.
CountryAustraliaUnited StatesNew ZealandUnited KingdomCanadaSingaporeOtherI'm an MSP, MSSP, or reseller
Δ
An autonomous SOC delegates selected investigation, closure, and response work to AI agents. Agents proceed within defined permissions, while people set authority, review results, handle exceptions, and remain accountable for operations.
Bounded tasks can run without approval at every step. A fully unattended SOC is a much broader proposition because uncertain evidence, business consequences, and novel incidents still need judgment and operational ownership. Organizations can gain substantial capacity without making every decision autonomous.
Human in the loop places approval before a defined decision or action. Human on the loop allows pre-authorized work to proceed under supervision, with people reviewing outcomes and able to intervene. Both require clear responsibilities and functioning escalation paths.
The organization and its designated owners remain accountable. Assign ownership for each workflow, define the provider’s responsibilities where applicable, and require approval before increasing authority. An AI verdict does not transfer business accountability to the model.
Use evidence requirements, target checks, least privilege, approval rules, execution limits, and independent review. Test recovery and workflow suspension, including what happens to requests already in progress. These controls reduce risk but cannot guarantee every action will be correct.
Yes. Attackers can try to influence investigations through content agents read or trigger automated responses through crafted activity. Treat evidence as untrusted, enforce permissions outside the model, limit action scope, and test both attempted manipulation and ordinary mistakes.
Autonomous Alert Triage
AI SOC Agents
AI SOC Analyst