EU Artificial Intelligence Act · Regulation (EU) 2024/1689

Article 14 compliance
built into your AI agents.

The EU AI Act requires human oversight of high-risk AI systems. Cheqpoint intercepts every AI agent action, routes it to a qualified human reviewer, and generates regulator-ready compliance evidence — automatically.

Article 9 — Risk ManagementArticle 12 — Record-keepingArticle 13 — TransparencyArticle 14 — Human OversightArticle 17 — Quality Management
Regulation (EU) 2024/1689

What Article 14 requires — and what it costs if you miss it

High-risk AI deployers face fines of up to €15 million or 3% of global turnover for non-compliance. Obligations kick in from August 2026 — the planning window is now.

Art. 14(4)(a)
Monitor AI operation

Designated humans must be able to monitor the AI system in real time and understand what it is doing. Cheqpoint provides an Approval Inbox showing every pending AI action with full context.

Art. 14(4)(b)
Detect and address deviation

Humans must be able to identify when the AI is behaving unexpectedly. Cheqpoint risk-scores every request and flags anomalies for specialist review before any action executes.

Art. 14(4)(c)
Override or intervene

Humans must be able to override, reject, or modify AI outputs. Cheqpoint gives every reviewer Approve, Decline, and Modify buttons — with an instant kill switch to pause any agent.

Art. 14(4)(d)
Prevent over-reliance

The AI must not pressure humans into rubber-stamping decisions. Cheqpoint requires deliberate action (reason code + confirmation) and never auto-submits based on AI recommendations.

Who it's for

Built for regulated industries

If your AI agents touch money, health, employment, or access to services — the Act applies.

Fintech & BankingAnnex III 5(b)
  • Credit limit changes and loan approvals
  • Refund and chargeback decisions
  • Account access and security changes
  • AML / fraud flag decisions
  • Transaction limit overrides
InsuranceAnnex III 5(c)
  • Claims approval and payout decisions
  • Underwriting acceptance / rejection
  • Policy cancellation or modification
  • Premium adjustment decisions
  • Fraud flag decisions on claims
Healthcare & MedTechAnnex III 2(a)
  • Clinical decision support sign-off
  • Prescription and treatment suggestions
  • Patient record access by AI agents
  • Triage priority assignments
  • Diagnostic output review
HR & RecruitmentAnnex III 4
  • Candidate shortlisting decisions
  • AI interview scoring review
  • Performance review flags
  • Promotion / termination recommendations
  • Background check processing
Legal & RegTechAnnex III 6
  • Contract execution decisions
  • Regulatory filing submissions
  • Sanctions screening overrides
  • Compliance flag resolutions
  • Data subject request processing
E-commerce & RetailGeneral deployer obligations
  • High-value refund approvals
  • Account fraud suspensions
  • Bulk pricing or discount changes
  • Supplier payment authorisation
  • Customer data deletion requests
How it works

Three lines of code. Article 14 on day one.

01
AI agent submits for review

Your AI agent calls the Cheqpoint API before executing any regulated action. Execution blocks immediately — nothing happens until a human decides.

# Python — blocks until human approves or declines
result = cp.request_approval(
    action="credit_limit_change",
    summary="Increase limit £5,000 → £8,000 for #C-00412",
    risk_score=0.68,
)
02
Human reviewer sees context and decides

Your compliance officer gets a Slack alert and opens the request. They see the action type, AI confidence score, risk level, full payload, and any relevant context. They Approve, Decline, or Modify — with a reason code.

03
Compliance evidence is recorded automatically

Every decision is written to the immutable audit log. At any time, generate an Article 14 Compliance Report showing exactly who reviewed what, when, with what response time — ready to share with regulators.

The compliance report

Regulator-ready evidence. Generated automatically.

When an auditor or regulator asks how you comply with Article 14, go to Compliance Report and click Print. In 30 seconds you have a dated, referenced document showing every human oversight decision made in any period.

  • Auto-generated Article 14 compliance statement
  • Total requests reviewed, human vs auto-decided
  • Approved / Declined / Modified breakdown
  • P50 and P95 response time metrics
  • Per-AI-Assistant and per-reviewer activity
  • Unique document reference number for your audit file
  • Print-optimised PDF export — no sidebar, no clutter
Generate your first report
Cheqpoint
Article 14 Compliance Report
EU Artificial Intelligence Act · Human Oversight Evidence
1,847
Requests Reviewed
94.2%
Human Decision Rate
4.3m
Avg Response Time
97.8%
SLA Compliance
Compliance Statement: Acme Financial Ltd has maintained continuous human oversight of all AI agent actions during 1 Apr – 30 Jun 2026, in accordance with Article 14 of Regulation (EU) 2024/1689. A total of 1,741 human decisions were made across 3 AI Assistants by 8 authorised reviewers.
CHQ-F7A291-20260630 · Acme Financial Ltd
Everything included

The complete compliance stack

Human oversight inbox

Real-time approval inbox with simple and technical views. Every action visible before it executes.

Kill switch per AI agent

Pause any AI Assistant instantly from the dashboard — no code change required. Takes effect on the next request.

Tamper-evident audit log

Every request, decision, actor, and timestamp recorded. CSV export for your audit file.

Compliance Report

Auto-generated Article 14 evidence document. Dated, referenced, and PDF-exportable for regulators.

SLA tracking

Set response time targets. Track P50 and P95 response times. SLA breaches logged automatically.

Reviewer roles & routing

Route high-risk decisions to specialist reviewers. OWNER / ADMIN / REVIEWER / VIEWER access controls.

Security & access controls

HMAC-signed sessions, 2FA (TOTP), IP allowlist, SOC 2-ready. EU data residency available on Enterprise.

Risk scoring & auto-rules

Risk score every request (0–1). Auto-approve low-risk, policy-compliant actions. Manual review for anything above threshold.

SDK for every stack

Python, Node.js, HTTP. LangChain, LangGraph, CrewAI, OpenAI Assistants, Claude, Vercel AI SDK.

Who uses Cheqpoint

Four buyers. One decision.

Head of Compliance / Chief Risk Officer
Regulators are asking how we oversee our AI. I need documented evidence that humans are in the loop.
Cheqpoint generates the evidence automatically. Export an Article 14 Compliance Report for any period in 30 seconds.
VP Engineering / Head of AI
We're shipping AI agents that take real actions. I need an oversight layer that doesn't break our velocity.
Three lines of SDK code. The approval inbox is handled. Your engineers keep shipping.
Head of Operations / Customer Support
AI is approving refunds and account changes. I can't lose visibility of what it's doing.
Every AI action lands in the Approval Inbox. Your ops team reviews and decides — nothing executes without their sign-off.
CFO / General Counsel
Non-compliance with the EU AI Act means fines up to €15M. I need a defensible audit trail.
Dated, referenced compliance reports and a tamper-evident audit log that can be presented to any regulator.
Get started today

Article 14 compliance.
Running in production today.

Start a 14-day Growth trial. Connect your first AI agent. Generate your first compliance report. No credit card required.

14-day free trial · No credit card · Cancel anytime · GDPR DPA available