102 DAYS UNTIL EU AI ACT DEADLINE

See what your AI
is actually doing

Connect your codebase. Get a compliance score in 60 seconds. Every finding explained in plain English. No terminal required.

Trusted by teams preparing for August 2026

51+ compliance checks
7 framework integrations
4 regulatory frameworks
Apache 2.0 open source core

Your compliance score, explained

This is what you see after connecting your repo. Every finding links to the specific EU AI Act article and tells you exactly what to fix.

console.airblackbox.ai/scan/acme-ai-agent
68
Score

acme-ai-agent

Last scanned 2 minutes ago · 47 files analyzed · 6 articles checked

Art. 9 Risk Mgmt
85%
Art. 10 Data Gov
78%
Art. 11 Tech Docs
60%
Art. 12 Logging
35%
Art. 14 Oversight
90%
Art. 15 Robustness
55%

Top Findings (click to expand)

High No audit trail for LLM calls Art. 12
What this means: Your AI agent makes calls to OpenAI but none of those calls are being logged in a way that's tamper-evident. If a regulator asks "what did your AI do on March 15th?", you can't answer that question.

How to fix it: Add the AIR trust layer (3 lines of code) or route calls through the AIR Gateway. This creates a signed, verifiable log of every AI action automatically.
High No human override mechanism detected Art. 14
What this means: Your AI can take actions (sending emails, making decisions) with no way for a human to intervene or override. The EU AI Act requires that humans can stop the system at any point.

How to fix it: Add air-gate to your pipeline. It pauses risky actions and routes them to a human via Slack before the AI executes them. Takes about 15 minutes to set up.
Medium Missing technical documentation Art. 11
What this means: Your project doesn't have a model card, data governance doc, or risk assessment. Auditors expect to see these documents alongside your code.

How to fix it: Run air-blackbox export to auto-generate documentation templates based on what the scanner found in your code. Fill in the blanks and you're covered.
Low No bias testing framework found Art. 10
What this means: The scanner didn't find any bias or fairness testing in your codebase. While not always required, having documented bias checks strengthens your compliance posture.

How to fix it: Add fairness testing for your model outputs. AIR Blackbox includes a bias scanner (air-blackbox scan-bias .) that checks for 6 common bias patterns.

Three steps. Sixty seconds.

No installation. No terminal. No CLI flags. Just connect and scan.

01
🔗

Connect Your Code

Link a GitHub repo (Pro/Team) or paste a code snippet (Free). We only need read access. Your code is scanned in memory and never stored.

02
🔍

Get Your Score

51+ checks run across 6 EU AI Act articles, plus ISO 42001, NIST AI RMF, and Colorado SB 24-205 crosswalks. Results in under 60 seconds.

03
📝

Read the Explanation

Every finding is explained in plain English. What it means for your business. What to fix. How long it takes. No jargon. No guesswork.

04
📄

Export the Report

Download a PDF report you can hand to your compliance team, board, or regulator. Includes remediation roadmap with time estimates.

You shouldn't need to be a developer
to know if your AI is compliant

CTO / VP Engineering

"The board wants an AI compliance update by Friday"

Connect your repos, get a score, export a PDF. Send it to the board with a remediation timeline. Done before lunch.

AI Governance Lead

"I need to audit 12 AI systems across 4 teams"

Scan every repo from one dashboard. Track compliance scores over time. Prioritize the highest-risk gaps first.

Recruiter / HR Tech

"We built an AI screening tool. Is it legal?"

AI in hiring is explicitly classified as high-risk under the EU AI Act. Scan your tool now. Know exactly where you stand before enforcement begins.

Compliance Manager

"I handle SOC2 and ISO but now they want AI too"

AIR Blackbox maps to ISO 42001, NIST AI RMF, and Colorado SB 24-205. Same audit structure you already know, applied to AI systems.

Start free. Scale when ready.

The open-source CLI is always free. Console adds the UI, GitHub integration, and plain-English explanations that make compliance accessible to your whole team.

Free
$0/month

Try the scanner. Paste code or upload a file. See what it finds.

  • 1 scan per month
  • Paste or upload code
  • Compliance score + top findings
  • Plain-English explanations
  • GitHub repo connect
  • Scan history
  • PDF report export
  • Team dashboard
Start Free Scan
Team
Custom

Tailored to your team size, repo count, and compliance needs.

  • Everything in Pro
  • Up to 5 team seats
  • Team compliance dashboard
  • Multi-repo scanning
  • Priority email support
  • Custom compliance policies
  • Quarterly compliance review call
  • Invoice billing
Contact Us

Annual plans: 2 months free. Pro $490/yr.

How it compares

AIR Console Enterprise Platforms Doc Generators
Scans actual code
Plain-English findings Jargon-heavy N/A
Scanning in 60 seconds Weeks to deploy
Post-quantum signatures ML-DSA-65
Open-source core Apache 2.0 Proprietary Some
Starting price $0/month $50,000+/year $0

Questions

Do you store my code?
No. Code is scanned in memory and discarded immediately. We store the scan results (scores and findings) so you can track progress, but never the source code itself. For paste/upload scans, the code is processed and deleted within seconds.
What GitHub permissions do you need?
Read-only access to repository contents. We use GitHub OAuth with the minimum required scope. You can revoke access anytime from your GitHub settings.
Is the open-source CLI going away?
Never. The CLI (pip install air-blackbox) is Apache 2.0 licensed and always will be. Console is a UI layer on top of the same scanner. If you prefer the terminal, nothing changes for you.
Does a passing score mean I'm legally compliant?
No. AIR Blackbox checks technical requirements from the EU AI Act. Legal compliance also requires organizational measures, documentation, and potentially conformity assessments. Think of the scanner as a starting point that identifies gaps, not a legal certification.
What languages and frameworks do you support?
The scanner analyzes Python AI projects and has specific integrations for LangChain, CrewAI, OpenAI SDK, Claude Agent SDK, Google ADK, AutoGen, and Haystack. It also detects common patterns in any Python AI code regardless of framework.
Can I scan private repos?
Yes. Pro and Team plans support private GitHub repos. After connecting via OAuth, you select which repos to scan. We only access repos you explicitly authorize.

Console launches June 2026

Get early access. First 100 signups get Pro free for 3 months.