Control before impact
Move from documenting AI risk after the event to controlling AI behaviour at the point where decisions are made and actions are triggered.
Explore runtime control →AI control before AI action.
OpterAI sits inside the execution path of enterprise AI systems. It evaluates prompts, model outputs, agent actions and workflow requests against policy, authority limits and business rules before execution.
Approve. Block. Escalate. Record. Before harm reaches the business.
Runtime controlDecision enforcement before execution.
Audit readyTraceable, explainable, defensible AI actions.
Regulated and high-stakes environmentsBuilt for industries where AI decisions carry consequences.
AI prompt or action requestCaptured from model, agent, workflow or SaaS tool.
InputPolicy and authority evaluationChecked against business rules, limits and governance conditions.
GateOpposition and verificationReasoning is challenged for truthfulness, bias, drift and risk.
ChallengeApprove, block or escalateUnsafe actions are stopped before they enter production.
DecisionAudit trail and evidence recordEvery event is timestamped, traceable and reviewable.
ProofStrategic shiftBefore
From post-incident reporting and static governance evidence.
Operational controlDuring
To policy-based permissioning at machine speed.
What makes OpterAI different
OpterAI is not a GRC platform.
Most AI governance tools tell you what happened after the AI acted. OpterAI controls whether the AI action is allowed to happen at all.
GRC systems document policies, risks and controls around systems. OpterAI operates inside the execution path of AI systems themselves. It controls whether an AI action is allowed to execute.
Traditional GRC
Documents the policy
Risk registers, attestations and audit evidence about what should happen.
OpterAI
Enforces the policy at runtime
Intercepts prompts and AI actions, applies constitutional controls and business rules, blocks unsafe execution, alerts reviewers and records every decision.
The execution problem
AI risk is no longer just a policy problem.
AI systems now generate decisions faster than conventional governance, risk and compliance processes can respond. AITru is designed for the moment where a decision is about to execute and the organisation needs a defensible answer: should this action be allowed to happen?
See the control modelPreventStop harmful decisions before operational damage occurs.
EscalateRoute material decisions to named human authority.
ProvePreserve the evidence needed for audit and review.
ScaleDeploy more AI without surrendering operational control.
Enterprise AI control infrastructure
A practical control layer for high-stakes AI.
OpterAI is AITru’s real-time constitutional control layer. It sits in the execution path of AI systems, evaluates each proposed action, and only permits execution when the action is safe, lawful, aligned and auditable.
Defined authority boundaries
Evaluate AI activity against organisational policy, ethical constraints, operational rules and human authority conditions.
View capabilities →Traceability by design
Create a structured evidence trail showing what was requested, assessed, challenged, approved, blocked or escalated.
Review regulatory outcomes →OpterAI
Always on. Always watching. Always ahead.
OpterAI gives enterprises a command centre for AI decisions across public LLM assistants like ChatGPT, Claude, Gemini and Copilot, internal models, agent networks, SaaS-embedded AI, automated workflows and external model APIs.
The control system for AI decisions
Before an AI action executes, OpterAI evaluates the request against policy, authority limits and governance rules. The decision can be approved, blocked, modified or escalated before risk propagates into the business.
- Validates AI decisions at the moment of execution.
- Blocks or escalates unsafe, non-compliant or unauthorised actions.
- Supports guided prompt compliance to help users remain productive.
- Records every governed AI interaction with operational traceability.
How OpterAI deploys. Coverage is defined by what is connected. Within the customer estate, OpterAI is configured against a set of connected systems (models, agents, SaaS tools, workflows and APIs). Together these form the control perimeter. Governance applies to active links inside that perimeter, not to systems outside it.
Constitutional protection system architecture
OpterAI operates as part of AITru’s constitutional protection system, translating rules, challenge, verification and evidence into real-time operational control.
Input capturePrompt, model, agent, workflow
Harm gateInitial safety and authority screening
Opposition networkChallenge, contradiction and reasoning review
TruthOps verificationValidation, provenance and integrity checks
Human authority controlApproval, escalation or stop-power
Audit evidenceTimestamped, traceable decision record
See it in action
OpterAI at the moment of execution.
Two short animations illustrating how OpterAI operates at runtime, followed by real screenshots from the live platform and a sample of the governance report it produces.
How OpterAI works, illustrated.
A short illustrative animation of OpterAI's runtime control flow, showing how prompts and AI actions are intercepted, evaluated against policy, and recorded into the audit trail. The two screenshots below are from the live platform itself.
An unsafe ChatGPT prompt, intercepted.
An illustration of how OpterAI handles an unsafe prompt directed at ChatGPT. The request is refused, the user is shown a constitutional verdict, and the event is sealed into the audit chain before the model can respond.
The operational console: real product screenshot.
A real screenshot from the live OpterAI platform. Every AI decision flowing through OpterAI is inspected, classified and surfaced in real time. Risk, compliance and security teams see what passed, what was warned, and what was blocked, without waiting for an end-of-quarter report or a manual log review.
84 AI decisions / 24h
20 blocked
15 warnings
49 passed
A real blocked prompt, captured with proof.
A real audit record from the OpterAI platform. When OpterAI blocks an action it doesn't just stop the request, it records it. The prompt, the constitutional verdict, the rationale shown back to the user, and the cryptographic record ID with previous-hash reference are all preserved together.
That hash chain is what turns a console event into auditable evidence. Each record is linked to the one before it, so tampering breaks the chain and is immediately detectable.
SHA-256 hash chain
Previous-hash linked
Timestamped UTC
Immutable record
For regulators & auditors
From live console to formal report.
Every event you see on the dashboard rolls up into a structured governance report: the same evidence pack regulators, internal audit and second-line risk teams receive.
Download a sample governance reportCore capabilities
Enterprise AI control without slowing the business.
The objective is not to produce more governance paperwork. The objective is to ensure AI decisions cannot execute unless they satisfy defined organisational, legal, ethical and operational controls.
Show all 8 capabilities
1
Real-time decision enforcement
Validate every governed AI action before execution and intervene when the request violates policy, ethics or authority boundaries.
2
Connected AI coverage
Govern AI activity across the configured control perimeter, including internal models, public LLM assistants, third-party SaaS tools, automation platforms and external APIs connected to OpterAI.
3
Runtime failure prevention
Stop hallucinations, misaligned optimisation, rogue execution and unauthorised actions before they cascade.
4
Complete audit trail
Capture the prompt, initiator, model, intended action, policy assessment, escalation path and final outcome.
5
Human authority enforcement
Ensure material decisions remain gated by defined human authority rather than self-authorised automation.
6
Prompt compliance guidance
Suggest safer or more compliant prompts where a user’s original request risks breaching policy.
7
Truth verification
Challenge reasoning and validate information to reduce hallucination, misinformation and synthetic contamination risk.
8
Board and auditor visibility
Give senior leaders, risk teams and auditors a live operational view of AI usage and control evidence.
Built around the enterprise, not generic AI policy
Your rules. Your reviewers. Enforced at machine speed.
Buyers do not just want “AI safety”. They want their own business rules enforced before AI actions create exposure, and they want named humans in the loop when material decisions arise.
Show how rules and human review work
Your rules, not generic AI policy
OpterAI enforces enterprise-specific rules at runtime, including:
- Decision authority and approval limits
- Financial commitment and exposure thresholds
- Customer-impact escalation rules
- Risk appetite boundaries
- Confidentiality and data-handling rules
- Regulatory and jurisdictional requirements
- Internal approval and sign-off workflows
Your policy library becomes an enforceable runtime control, not a slide deck attestation.
Human review is not an afterthought
When OpterAI detects a material, unsafe or policy-sensitive AI action, it operates an end-to-end review workflow:
- Alerts the designated reviewer mapped to the breached policy
- Holds the AI action before it can execute
- Records the escalation with full decision context
- Updates the OpterAI Control Panel with the reviewer’s decision
- Captures remedial action for audit and continuous improvement
Human oversight as operational workflow, not a slide in a governance deck.
Where OpterAI wins
Concrete answers to what our customers actually ask.
Here is exactly what OpterAI's runtime constitutional control looks like in practice, and why other platforms cannot deliver it.
Show answers & platform comparison
01
What OpterAI stops
- Prompts containing customer PII, PHI, PCI or confidential pricing leaving the perimeter to ChatGPT, Claude, Gemini or Copilot.
- Agent tool-calls that would write to production systems without the required human approval.
- AI-generated decisions that exceed defined authority limits, financial, contractual, operational or customer-facing commitments.
- Model outputs that fail truth, bias or jurisdictional checks before they reach a customer.
- Multi-step agent workflows where any single step breaches policy, even when each step looks safe in isolation.
02
Who gets alerted
- The user, with a real-time block, a plain-English reason, and a compliant alternative prompt where one exists.
- The named human authority, the specific approver mapped to the policy that was breached, routed via Slack, Teams, email or webhook.
- Risk, compliance and audit, through a live operational console and rule-based notification routing.
- Security and SOC, when a block is also a security signal: data exfiltration attempt, prompt injection, jailbreak.
- Board and regulators, via scheduled or on-demand evidence packs generated from the audit record.
03
What evidence is created
- The exact prompt, agent call or model output that triggered the event.
- The identity of the initiator, human or machine, and the system of origin.
- The policy version, rule and clause that was evaluated, with the decision rationale.
- The authority path taken, approve, block, modify or escalate, with timestamps at every step.
- An immutable, timestamped audit record structured for regulator submission and external assurance.
Adjacent tools
Why audit and GRC platforms cannot deliver this?
These platforms are valuable, but each is built for a different layer of the problem. OpterAI is the layer that decides, in real time, whether an AI action is permitted to execute.
Purview, Defender, Copilot Studio
Designed for: data classification, identity, endpoint protection and Copilot configuration inside the Microsoft 365 estate.
Where it ends: these tools govern data access and tenant boundaries. They do not evaluate the reasoning, intent or downstream action of an AI decision, and they do not extend across non-Microsoft models, agents and SaaS.
OpterAI differenceConstitutional evaluation of the decision itself, across every model, agent and tool, not just the Microsoft stack.
AI Control Tower & Now Assist governance
Designed for: orchestrating workflows, tickets and risk registers around AI usage inside the ServiceNow platform.
Where it ends: it documents, routes and reports. It does not sit in the execution path of an AI action and block it at machine speed before damage occurs.
OpterAI differencePre-execution permissioning inline with the AI call, not a workflow that runs after the fact.
Archer, OneTrust, MetricStream, ServiceNow GRC
Designed for: policy management, control libraries, risk registers and attestations for the enterprise compliance programme.
Where it ends: they hold the rules and the evidence of attestation. They do not enforce a rule against a live AI decision before it executes.
OpterAI differenceTranslates the policy held in the GRC system into an enforceable runtime control, the missing actuator between policy and execution.
Audit platforms, SIEM, AI observability tools
Designed for: logging, monitoring, drift detection and post-hoc forensics across AI and security activity.
Where it ends: they tell you something went wrong, after the unsafe AI action has already executed and the data has already moved.
OpterAI differencePrevention before execution, with the evidence record produced as a by-product of the block, not as the only output.
Why runtime control matters
From compliance reporting to compliance enforcement.
Traditional governance systems document policies and risk. OpterAI is designed to control whether AI actions are allowed to proceed in the first place.
Post-incident oversight
- Compliance validated after execution.
- Limited visibility into AI reasoning and intent.
- Fragmented or incomplete decision evidence.
- Regulatory exposure discovered after damage.
- Human oversight too slow for machine-speed action.
Pre-execution permissioning
- Decisions blocked when they breach policy or authority.
- Full decision lineage and operational proof.
- Continuous audit readiness across AI estates.
- Human escalation before irreversible action.
- AI operates within enforceable control boundaries.
Markets we serve
Built for industries where AI decisions carry consequences.
OpterAI is engineered for environments where decisions affect customers, capital, patients, networks, regulation, reputation and legal accountability.
Show 4 industries
Insurance & reinsurance
Real-time control for underwriting, claims, pricing, tail-risk assessment and AI-driven risk decisions.
Show reinsurance proof
Reinsurance proof
Built for the decisions that move capital.
Reinsurance runs on treaty terms, delegated authority and exposure limits. When AI is involved in pricing, capacity allocation, claims and tail-risk assessment, the same rules still apply, and they have to apply at the moment the AI action is about to execute.
Where OpterAI sits in a reinsurance workflow
OpterAI is configured against the connected systems inside the cedent or reinsurer estate: pricing models, underwriting copilots, claims agents, broker-facing assistants and downstream SaaS tools. Together these define the control perimeter. Every governed AI action inside that perimeter is evaluated against the firm’s own treaty, authority and risk rules before it executes.
- Pre-execution checks on AI-generated pricing, terms and quotes.
- Authority routing when an AI action would commit capacity above a defined limit.
- Escalation when a claims or coverage decision touches a sensitive class of business.
- An evidence record of every approved, blocked and escalated AI decision.
Concrete business rule
An example of the kind of rule OpterAI enforces at runtime, expressed in plain reinsurance language.
RuleTreaty limit and delegated authority check
TriggerAI pricing or quote action on a treaty placement
CheckAggregate exposure plus quoted line versus treaty limit and underwriter authority
ThresholdRisk threshold for cat-exposed lines or sanctioned territories
OutcomeApprove within authority, escalate to the named approver, or block
EvidenceTimestamped record of the decision, rule applied and authority path
AI quotes inside authority
Blocks AI-generated quotes that exceed treaty limits, delegated authority or risk appetite, and routes them to the named underwriter before they reach the broker.
Sensitive decisions stay with humans
Escalates AI-recommended claims decisions on large losses, contested coverage or sanctioned parties to the right adjuster or claims lead, with the reasoning preserved.
Tail risk under control
Applies risk-threshold checks to AI-driven accumulation, pricing and capacity decisions on cat-exposed or correlated lines, and creates evidence for the actuarial and risk function.
Banking & financial services
Govern credit, trading, compliance and customer decisions with auditability and authority control.
Healthcare & pharmaceuticals
Support patient safety, clinical accuracy and accountability across AI-assisted clinical decisions, drug discovery, clinical trial design, pharmacovigilance, regulatory submissions and GxP-compliant manufacturing.
Telecommunications
Govern network, automation and customer-facing AI with real-time reliability and accountability.
Regulatory outcomes
Designed to support explainability, traceability and human oversight.
AITru does not replace compliance frameworks. It supports them by helping translate governance expectations into runtime controls and decision records, aligned to the major global AI standards and sector-specific regulation.
Show all 7 standards & frameworks
EU AI Act readiness
Supports the transparency, human oversight, accuracy and risk-management expectations applicable to high-impact AI activity under the EU AI Act, and creates evidence for the controls in place.
NIST AI Risk Management Framework
Aligns to the NIST AI RMF functions of Govern, Map, Measure and Manage, with runtime controls and evidence to support continuous monitoring.
ISO/IEC 42001 alignment
Aligns to ISO/IEC 42001 control objectives and creates evidence for organisations establishing, maintaining and improving an AI management system.
UK AI principles & sector regulators
Supports the UK’s pro-innovation AI principles and regulator expectations across the FCA, PRA, ICO, Ofcom and MHRA.
Model governance and resilience
Supports expectations around explainability, accountability, model governance, data integrity and operational resilience (SR 11-7, DORA, PRA SS1/23).
Pharmaceutical AI regulation
Aligns to FDA AI/ML guidance, EMA’s reflection paper on AI in the medicinal product lifecycle, MHRA AI Airlock, GxP standards (GMP, GLP, GCP, GVP), 21 CFR Part 11, EU Annex 11, GAMP 5 computer system validation and ALCOA+ data integrity, and creates evidence for AI activity across drug discovery, clinical trials, pharmacovigilance and manufacturing.
OECD, G7 Hiroshima & Singapore Model AI
Creates evidence for validity, reliability, auditability, fairness, human-centric governance and responsible deployment, aligned to multiple international frameworks.
Latest news and white papers
Resources for enterprise AI control.
Explore AITru’s current papers and product material on OpterAI, AI runtime constitutional control and the core capabilities that differentiate it from adjacent governance tools.
Show 8 papers & resources
OpterAI overview
A concise two-page overview of OpterAI and its real-time AI control proposition.
Request paper →Runtime decision control for agentic AI
The ten enterprise risks agentic AI introduces — goal manipulation, decision drift, authority escalation, memory poisoning and more — and why runtime decision control is the new control layer for AI-influenced action.
Download PDF →Insurance AI has moved beyond pilots
How insurance has crossed from AI experimentation into controlled execution — and why the next competitive advantage is runtime AI control, not more pilots.
Download PDF →AI in pharmaceuticals
Why pharma's strategic question has shifted from accelerating AI to proving runtime control across discovery, clinical trials, pharmacovigilance, regulatory writing and GxP manufacturing.
Download PDF →The New Tail: AI, Model Drift & Catastrophe Risk in Reinsurance
How AI-influenced systems compress uncertainty, drift catastrophe model outputs and create correlated risk across underwriting, exposure and capital — and why reinsurers need runtime control before uncertainty becomes exposure, before exposure becomes capital strain.
Download PDF →OpterAI and AI
Outlines the case for AI runtime constitutional control in reinsurance and regulated enterprise environments.
Request paper →Top 10 OpterAI differentiators
Explains the key differences between OpterAI and conventional AI governance or monitoring approaches.
Request paper →OpterAI governance report
A real, downloadable evidence pack: the format regulators, auditors and second-line risk teams receive. Reference IDs, timestamps, severity and source for every AI decision.
Download PDF →Get in touch
Book an AI control briefing.
A 30-minute runtime AI control briefing for leadership, risk, compliance and AI engineering teams. See exactly how OpterAI intercepts, evaluates and records AI activity inside your execution path, and how it would apply to your highest-stakes AI use cases.