Frequently Asked Questions

Regiment AI is Infopercept's governed AI security command platform. It unifies offensive security (AI pentesting), defensive operations, and compliance into one operating model, so every AI-driven action is bounded, approved, and auditable instead of running unsupervised. It's built by people who have worked the full attack lifecycle: red team, SOC, and compliance.

Neither. Regiment AI runs specialized agents - in three battalions - that reason, investigate, and act inside explicit rules of engagement. Simple, repeatable work runs deterministically. AI reasoning is reserved for ambiguous judgment calls: hunting exposure, modeling attack paths, decoding a novel vulnerability. High-consequence actions route to a human for approval before they execute.

Most enterprises don't have a tooling problem - they have an execution problem. Critical decisions cross too many systems, teams, and approval boundaries before risk becomes action, and evidence reconstruction for leadership or an audit is slow and costly. Regiment AI creates one accountable operating model across the estate you already own, without a rip-and-replace or an uncontrolled leap into agent autonomy.

Measurable impact. Regiment AI's impact is measured on four outcomes:

• Faster time from risk to action: up to 99% less time from an alert or finding to a governed decision or fix, compared with manual handoffs across teams and approvals.
• Evidence prepared automatically: 10+ hours saved per audit or leadership review, because every action is logged with its policy, approval and result as it happens. Nobody has to piece the evidence together afterwards.
• Analyst time saved: 5+ hours saved per analyst per day, because alert triage and investigation run automatically and analysts only review the cases that need a human decision.
• Faster offensive testing: an AI pentest is completed in 3-7 hours instead of ~5 days of manual effort, so assets can be tested continuously instead of once or twice a year.

Security teams that need continuous validation of their own defenses, consistent AI-assisted detection and response, and audit-ready compliance evidence - without handing an AI agent unrestricted autonomy over production systems.

It triages alerts, enriches detections, and summarizes cases using pre-built recipes. Every workflow is policy-gated, and high-risk actions require explicit human approval before they execute. An immutable audit log is written on every step.

It runs continuous, AI-driven penetration testing: discovering exposure, chaining attack paths, and validating real exploitability, rather than waiting for a scheduled pen test. Hard technical guardrails prevent out-of-scope reach, every offensive action needs sign-off from two authorized humans, and every finding ships with proof and business-impact context.

It continuously verifies controls against frameworks like SOC 2 and ISO 27001, and produces audit-ready evidence packages on demand - so compliance is a byproduct of daily security work rather than a periodic scramble, with no ambiguity between what a control intends and what's actually in place.

Regiment AI's two remediation agents. Remedy X contains and fixes active threats and exploitable misconfigurations the moment they're confirmed. Remedy Y reverts risky code or configuration drift back to an approved baseline. Both act only inside pre-approved playbooks, and every fix is scoped, logged, and reversible - never silent.

A control called Scoped Authority. Every agent operates inside an explicit boundary of assets, actions, and data set by your team. Anything outside that declared scope is unreachable, regardless of what the agent's reasoning suggests.

Dual approval requires sign-off from two authorized humans before a high-risk or offensive action executes - for example, before Scanners & Strikers validates an exploit path, or before Remedy X/Y makes an autonomous fix. It removes any single point of autonomous failure.

Yes. Any authorized user can trigger the kill switch to halt an agent's execution instantly, without waiting on a support ticket.

Every decision, approval, override, and outcome is written to an immutable audit log. Entries can be reviewed but never edited, giving auditors and leadership a defensible record from decision through outcome.

Yes. Regiment AI operates inside your existing RBAC policies. An agent acting on behalf of a user cannot reach modules, data, or capabilities that user isn't already authorized to view.

Deterministic execution is the default for repeatable, well-understood tasks - playbooks, policy enforcement, machine-speed response. AI reasoning is reserved for the unknowns: correlating signals, prioritizing exposure, chaining an exploit path. Once an AI-found pattern is validated, it's codified into deterministic automation, so the same pattern doesn't require AI reasoning the next time.

Your data is never used to train AI models - not Regiment AI's, not the model provider's, and not AWS's.

Every AI request is protected at two layers:

• Invinsense AI Firewall: Every prompt passes through Infopercept's Invinsense AI Firewall before it reaches a model. The firewall applies guardrails and encrypts sensitive information in the prompt, so sensitive values from your data are never exposed to the model in the clear.
• Regional Amazon Bedrock inference: Models run on Amazon Bedrock in your chosen AWS region, so inference stays within that regional boundary. Under AWS's written commitment, prompts and responses are not shared with the model provider, are not stored or logged by Bedrock, and are not used to train any model.

Request early access. Infopercept's team scopes your environment - the assets in play, your existing tools, and which battalion matters most on day one - before deployment begins.

An inventory of the assets and identity systems you want defended and validated, the SIEM/XDR/EDR/NDR tooling already in place, and internal agreement on your starting rules of engagement - who approves offensive actions, and who holds kill-switch authority.

Regiment AI is built to deploy in weeks, not months. Intelligence compounds across the three battalions from day one as it connects into your existing estate.

On defense: your existing SIEM, XDR, EDR, and NDR tooling. On offense: your exposure and attack-surface tooling - AI pentest, BAS, CART, ASM, and VM. Every decision routes through one shared policy engine regardless of which tool produced the signal.

No. Regiment AI is designed to sit across your current estate as a governed decision layer, not a rip-and-replace platform.

Yes - you have both options.

• Infopercept-provided inference: Regiment AI comes with managed inference out of the box. Models run on regional Amazon Bedrock, and every prompt passes through the Invinsense AI Firewall. There is nothing to set up on your side.
• Bring your own inference: if you already operate AI infrastructure - your own cloud model account or self-hosted models - you can connect it to Regiment AI, and inference stays inside your environment. Prompts still pass through the Invinsense AI Firewall, so the same guardrails and protection of sensitive information apply.

Either way, the same policy engine, scoping, and audit trail govern every action. The choice changes where inference happens, not how it is controlled.

Traditional MDR is largely alert triage and escalation from Tier 1/Tier 2 analysts, with offensive validation rarely included and compliance evidence collected periodically. Regiment AI runs a continuous, closed loop from offensive validation into defensive detection into compliance evidence, with human review only where a decision carries real consequences.

Most AI SOC tools add investigation assistance on top of a stack that still carries the full workload, and lean on generative AI for nearly everything. Regiment AI uses AI selectively, for judgment calls only; keeps everything else deterministic and fully auditable; and treats offense, defense, and compliance as one operating model instead of three disconnected tools.

Regiment AI is in early access. Infopercept is onboarding select security teams ahead of general availability.

Talk to the team directly - book a meeting to start the conversation and scope your deployment.