AIUC-1 covers 6 domains — data privacy, security, safety, reliability, accountability, and societal impact. Aguardic implements its 49 requirements as 64 enforceable rules that run automatically on every AI output, code commit, and document.
14-day free trial · No credit card · Free AIUC-1 policy pack
81%
Score
8
Violations
4
Open
6/6
Policies
Policy Coverage
Requirements Coverage
AIUC-1 defines 49 active requirements across 6 domains. Aguardic implements them as 64 granular rules. The table below shows representative requirements per domain, what Aguardic enforces, the evidence it produces, and the work your team still owns.
A003
Restrict AI agent data collection to what is necessary for the task at hand.
How Aguardic helps
Input filtering policies limit what data AI agents can access and collect. Policies enforce data minimization across every connected integration.
Evidence produced
Data collection policy logs · blocked access records
What you handle
Define which data fields each agent legitimately needs for its task.
A006
Prevent personally identifiable information from being exposed in AI outputs.
How Aguardic helps
PII detection policies scan AI outputs, documents, emails, and messages. Block or redact PII before it reaches unauthorized recipients.
Evidence produced
PII detection logs · blocked output records
What you handle
Define your organization's PII classification (what counts, what's allowed where) and approve redaction rules.
A001
Define and enforce policies governing data provided to AI systems.
How Aguardic helps
Input filtering policies govern data entering AI systems. Full data classification taxonomy requires manual definition.
Evidence produced
Input policy evaluation logs
What you handle
Define your data classification taxonomy (public / internal / confidential / restricted) that Aguardic enforces against.
B002
Detect and mitigate adversarial inputs designed to manipulate AI behavior.
How Aguardic helps
Semantic AI evaluation detects prompt injection, jailbreaking, and adversarial inputs. Deterministic rules catch known attack patterns.
Evidence produced
Adversarial detection logs · blocked input records
What you handle
Review novel adversarial patterns surfaced by Aguardic and decide how to update detection rules.
B006
Ensure AI agents only perform authorized actions within defined boundaries.
How Aguardic helps
Agent behavior policies restrict tool calls, API access, and actions to authorized boundaries. Unauthorized actions are blocked before execution.
Evidence produced
Agent action logs · blocked tool-call records
What you handle
Author the allow-list of tool calls and approved actions per agent.
B001
Conduct third-party adversarial testing of AI systems at defined intervals.
How Aguardic helps
Aguardic provides continuous automated testing but does not replace independent red-team assessments.
What you handle
Engage a red-team vendor on a defined cadence. Aguardic's continuous testing complements but doesn't replace third-party assessments.
C003
Prevent AI systems from generating harmful, dangerous, or toxic outputs.
How Aguardic helps
Output scanning policies detect and block harmful content including toxicity, dangerous instructions, and inappropriate material across every surface.
Evidence produced
Harmful content detection logs · blocked output records
What you handle
Define your org's safety policy (what's acceptable vs blocked) and approve enforcement thresholds.
C007
Identify and flag AI outputs that carry elevated risk for human review.
How Aguardic helps
Policy severity levels flag high-risk outputs for human review. Warn and escalate modes ensure flagged outputs receive attention.
Evidence produced
Flagged output logs · escalation records
What you handle
Staff the human review queue and set severity thresholds that require review.
C001
Define a comprehensive taxonomy of risks for AI system outputs.
How Aguardic helps
Aguardic enforces policies based on defined taxonomies but does not create the taxonomy itself.
What you handle
Author your domain-specific risk taxonomy — what categories of harm matter for your product and users.
D001
Detect and prevent AI systems from generating factually incorrect information.
How Aguardic helps
Knowledge RAG cross-references outputs against your knowledge bases. Semantic AI evaluates factual consistency. Deterministic rules catch known hallucination patterns.
Evidence produced
Hallucination detection logs · knowledge verification records
What you handle
Upload your authoritative knowledge sources to the knowledge base and keep them current.
D003
Prevent AI agents from making unsafe or unauthorized tool calls.
How Aguardic helps
Agent behavior policies validate tool calls against allow-lists and parameter constraints. Unsafe calls are blocked before execution.
Evidence produced
Tool call validation logs · blocked action records
What you handle
Author the allow-list and parameter constraints per tool exposed to agents.
E015
Maintain comprehensive logs of all AI model activity for audit purposes.
How Aguardic helps
Every policy evaluation is logged with timestamp, input, result, violations, and trace ID. Every AI interaction is logged and exportable.
Evidence produced
Exportable evaluation logs · model activity records · decision-point audit trail
What you handle
Define retention policies and integrate Aguardic log exports with your SIEM or evidence repository.
E016
Implement mechanisms to disclose when AI is being used in interactions.
How Aguardic helps
AI disclosure policies enforce transparency requirements. Policies can require AI-generated content labeling.
Evidence produced
Disclosure enforcement logs · labeling compliance records
What you handle
Approve the disclosure copy your users see; Aguardic enforces that the disclosure is present.
E010
Enforce organizational acceptable use policies for AI systems.
How Aguardic helps
Custom policies encode acceptable use rules. Enforcement across connected integrations ensures organization-wide compliance.
Evidence produced
Acceptable use policy logs · violation records
What you handle
Author the organizational AUP; Aguardic enforces the rules you define.
E001–E003
Establish, test, and maintain incident response plans for AI-related incidents.
How Aguardic helps
Aguardic provides automated first-response but does not substitute for a documented, tested incident response program.
What you handle
Draft, test, and maintain the AI-specific incident response plan and tabletop exercises.
F001
Prevent AI systems from being used for cyberattacks or malicious purposes.
How Aguardic helps
Policies detect and block attempts to use AI for code exploitation, social engineering, and other cyber misuse. Real-time enforcement prevents misuse.
Evidence produced
Misuse detection logs · blocked attempt records
What you handle
Define what counts as misuse for your product context and keep detection rules updated.
F002
Prevent AI systems from being used for catastrophic harm.
How Aguardic helps
Critical severity policies immediately block outputs related to weapons, CBRN threats, and catastrophic harm. Zero-tolerance enforcement.
Evidence produced
Critical violation logs · blocked catastrophic content records
What you handle
Define catastrophic-misuse categories for your context (weapons, CBRN, etc.) and approve zero-tolerance thresholds.
Coverage mappings reflect Aguardic's current product capabilities mapped to representative AIUC-1 requirements. The full AIUC-1 pack covers all 49 active requirements via 64 granular rules. Validate these mappings for your specific deployment context.
The 6 AIUC-1 Domains
14 rules
Input and output data policies, limiting AI agent data collection, protecting IP and trade secrets, preventing cross-customer data exposure, PII leakage prevention, and IP violation prevention.
View in Marketplace13 rules
Prompt injection defense, jailbreak detection, encoded payload blocking, scraping prevention, harmful input filtering, agent scope enforcement, unauthorized tool blocking, and output over-exposure limits.
View in Marketplace14 rules
Harmful output prevention, hostile and discriminatory content blocking, deceptive content detection, high-risk advice disclaimers, out-of-scope output prevention, output vulnerability scanning (SQLi, XSS, command injection), and risk monitoring.
View in Marketplace10 rules
Hallucination prevention, fabricated citation and statistic detection, missing uncertainty flagging, unsafe tool call restrictions, privilege escalation detection, and operating schedule enforcement.
View in Marketplace7 rules
Acceptable use policy enforcement, data extraction prevention, misrepresentation detection, audit trail completeness, AI disclosure in external communications, and autonomous action transparency.
View in Marketplace6 rules
Malware generation blocking, attack planning prevention, vulnerability exploitation detection, CBRN instruction blocking, mass harm prevention, and dual-use research safeguards.
View in MarketplaceAll 6 templates available as a free pack in the AIUC-1 marketplace
How It Works
One-click install from the marketplace. 6 templates covering all 64 rules across all AIUC-1 domains — deterministic regex patterns for speed, semantic LLM analysis for nuance.
Link OpenAI, Anthropic, Gemini, GitHub, Slack, and 10+ more integrations. Every AI output, code commit, and document is evaluated against AIUC-1 rules automatically.
Violations are caught in real-time with full audit trails. Each evaluation maps to specific AIUC-1 control IDs — export evidence for auditors, regulators, or internal review.
Already have internal AI governance documents? Upload them and extract enforceable rules automatically
Cross-Framework Coverage
AIUC-1 is comprehensive by design. Its 6 domains overlap significantly with other major AI governance frameworks — meaning one AIUC-1 implementation gives you a head start on multiple compliance requirements.
Significant overlap
Annex A controls (A.7.2, A.7.3, A.8.4, A.9.3, A.9.4, A.10.2) map to AIUC-1 data privacy, safety, and accountability domains.
Significant overlap
Articles 9, 11, 13, 14, 15, and 52 map to AIUC-1 across data privacy, safety, accountability, and transparency requirements.
Significant overlap
NIST's Map, Measure, Manage, and Govern functions correspond to AIUC-1's safety, reliability, and accountability domains.
Partial overlap
AI-specific security risks (model poisoning, prompt injection, data leakage) covered by AIUC-1 Domain B security controls.
Partial overlap
Cloud-native AI controls for data governance, model lifecycle, and operational security align with AIUC-1 Domains A and B.
Implement AIUC-1 once — get a head start on 5 major AI governance frameworks.
Audit Ready
Every evaluation result maps to a specific AIUC-1 control ID (A003, B002, C003, etc.). Export audit packages grouped by domain for streamlined reviews.
No more point-in-time assessments. Every AI output, code commit, and document is evaluated against AIUC-1 rules in real-time — evidence generates itself.
AIUC-1 is updated quarterly by the AIUC consortium. When the standard is updated, Aguardic updates the policy pack to match. Subscribers get new rules without manual intervention.
Frequently Asked Questions
AIUC-1 is the first security, safety, and reliability standard built specifically for AI agents. Created by the Artificial Intelligence Underwriting Company (AIUC) — a consortium of 100+ Fortune 500 CISOs with technical contributors from Cisco, MITRE, Stanford, Microsoft, and Anthropic — it defines 49 active requirements across 6 domains. Certification is independently audited by Schellman.
Yes. The complete AIUC-1 policy pack is free on the Aguardic Marketplace. You can install all 6 domain templates with one click and start enforcing immediately. No subscription required for the policy pack itself.
The pack includes 64 enforceable rules across 6 templates — one for each AIUC-1 domain. Rules use a combination of deterministic evaluation (regex pattern matching for speed) and semantic evaluation (LLM analysis for nuanced checks).
Enforced requirements (23 of 49) can be validated technically — implemented as 64 granular rules covering output scanning, input filtering, agent behavior monitoring, and configuration checks. Documented requirements (26 of 49) are procedural (incident response plans, vendor due diligence, risk taxonomy definitions) that Aguardic tracks and reports on but cannot automatically validate.
Yes, significantly. AIUC-1 publishes official crosswalks to ISO 42001, EU AI Act, NIST AI RMF, OWASP Top 10, and CSA AICM. Each AIUC-1 requirement page on aiuc-1.com shows the specific control mappings to these frameworks.
Sign up for Aguardic (free trial available), navigate to the Marketplace, find the AIUC-1 category, and click Install on any or all 6 domain templates. Connect your AI tools and enforcement starts automatically.
Yes. When the AIUC consortium publishes updates to AIUC-1 or releases AIUC-2, Aguardic will update the policy pack. Subscribers receive updates automatically — no manual intervention required.
AIUC-1 vendor assessment?
Upload it. We draft answers citing the AIUC-1 control catalog — governance, data handling, model lifecycle, transparency — describing what Aguardic enforces in production, not just what your governance docs assert.
Install the free AIUC-1 policy pack, connect your AI tools, and start generating audit-ready evidence in minutes.
Or explore the documentation
