Andres Jimenez Nanhekhan
I help organizations deploy impactful technology in a governed and safe environment.
About the project
Making the EU AI Act's quality management requirements easier to check
Providers of high-risk AI systems must run a quality management system under Article 17 of the EU AI Act. BS EN 18286:2026 (Artificial intelligence. Quality management system for EU AI Act regulatory purposes) is the European standard that sets out what that system needs to contain. It's detailed, and for many teams the first hard question is simply: where do we stand?
This tool gives organisations a structured, private way to answer that question — and a gap report to take to the next conversation with their team, their auditor or their board.
It was built by a team of three.
How the assessment works
Applicability first
We start by asking whether you provide or deploy AI. The requirements are mainly provider obligations, so a deployer isn't marked down for activities its provider controls. Each area also checks whether a process exists before going into detail.
Private by design
There is no server and no database. Your answers stay in your browser; a share link carries the results inside the link itself.
The AI Safety Indicator
Alongside the readiness results, the report shows an AI Safety Indicator: an awareness tool that suggests which of seven AI safety topics — bias and discrimination, incorrect outputs, robustness and drift, misuse, privacy and data leakage, data poisoning and supply-chain integrity, and specification gaming — may be less addressed by your current processes. It adds no questions; it is derived from the answers you have already given.
Each relevant question or option is linked to one or more topics with a weight: 1.0 where it directly targets the topic, 0.6 where it partly addresses it, and 0.3 where it helps indirectly. Answers are combined into one of four indicative levels per topic; no numbers are shown, "not sure" answers are listed as unknowns rather than counted, and provider-only activities are left out for deployers. The weights are a documented judgement call by our team, not a validated method, and the indicator does not measure how risky an AI system is.
The team
Anton Tugushev
PhD. AI governance professional with a background in international law, public policy, and technology strategy, combining experience at the World Bank and in public-sector digital transformation.
Camilla Balbis
Operations generalist focused on AI safety, trained in cybersecurity and ISO/IEC 42001 (Lead Auditor), and embedded in the AI safety field through AIGS Canada, BlueDot and ML4Good.
Limitations
This is an indicative self-assessment covering selected requirements of BS EN 18286:2026. It is not a conformity assessment, a certification, an audit or legal advice, and a good result here does not mean a system complies with the EU AI Act. It does not yet cover the additional requirements for AI systems that continue to learn after release (§ 8.8). Clause references follow the final draft text of the standard and should be checked against the published edition. The questions reflect our reading of the standard; they don't reproduce it, and the standard itself remains the authoritative source.