25-question self-check
How ready is your AI governance?
Score how you run AI across six areas, and get a list of gaps tagged by who fixes them: policy, a tool, or engineering. Built for SaaS teams that enterprise buyers will question.
Nothing you enter is stored or sent anywhere. This is a quick self-assessment, not an audit, a certificate or legal advice. See the questions first.
What the check asks
Three short questions about your situation, then 25 on how you run AI. Each answer is Yes, Partly, No or Not sure. Not sure counts as No.
About you
- Do you sell to customers in the EU, or does your AI's output reach people in the EU?
- Do you handle personal data about people in India?
- Do banks, NBFCs or other RBI-regulated firms buy from you?
Where do you stand?
- Inventory
- Do you keep a current list of every AI feature in your product and every AI tool your staff use? For each AI feature, do you know which model, vendor and version powers it? Do you know which customer data reaches each model or AI vendor? Have you looked for AI tools your staff use that nobody approved, using sign-in records, expenses or browser extensions?
- Policy and ownership
- Do you have a written AI use policy that staff have read and acknowledged? Is one named person accountable for AI risk, with the authority to pause or switch off an AI feature? Before a new AI feature ships, do you assess its risks and keep that record? Have the people who build, buy or use AI been trained on what it can get wrong?
- Data and vendors
- Do your contracts with AI vendors say your customers' data is not used to train their models? Do you know where each AI vendor processes and stores your data, how long it keeps it, and how it deletes it? Do you keep a list of your AI vendors' subprocessors, and hear when they change? Can you tell a customer which AI vendors touch their data, and can they switch AI features off for their account? Do you remove or mask personal data before it goes to a model, where the model doesn't need it?
- Testing and oversight
- Before release, do you test AI outputs against a written set of cases? Do you test for wrong, biased or harmful answers, and keep the results? Do you re-test when the vendor changes a model version or you change a prompt? Where AI affects decisions about people, such as hiring, credit or health, does a human review before action? Do users know when they are dealing with AI or reading AI-generated content?
- Security and logging
- Do you log AI requests and outputs, with access controls, so you can trace what happened? Have you tested your AI features for prompt injection and for data leaking between customers? Are AI API keys and agent permissions limited to what is needed, and rotated? Does an AI feature only retrieve data that the signed-in user is allowed to see?
- Incidents and evidence
- Do you have a written plan for an AI incident, such as a harmful output, a data leak through a model or a vendor outage, with who decides and who tells customers? Does that plan name the reporting deadlines that apply to you, such as CERT-In's six hours and, from 13 May 2027, the DPDP report to the Data Protection Board within 72 hours? Could you answer an AI security questionnaire from documents you already have: policy, register, test records and vendor terms?
What the check follows, and what it doesn't claim
The six areas follow the shape of the NIST AI RMF 1.0 functions (Govern, Map, Measure, Manage) and the Annex A controls of ISO/IEC 42001:2023. NIST describes the framework as voluntary and says a revision is in progress. Our NIST AI RMF guide for SaaS maps each function to the records to keep. The standard is a management-system standard, and only accredited certification bodies issue certificates against it. See ISO 42001 readiness if that is your route.
Closest match, by area
- Inventory
- NIST AI RMF Govern (GOVERN 1.6, the AI inventory) and Map. ISO/IEC 42001 Annex A.4, resources for AI systems.
- Policy and ownership
- NIST AI RMF Govern. ISO/IEC 42001 Annex A.2 policies, A.3 internal organization and A.5 impact assessment.
- Data and vendors
- NIST AI RMF Govern and Map. ISO/IEC 42001 Annex A.7 data for AI systems and A.10 third-party and customer relationships.
- Testing and oversight
- NIST AI RMF Measure. ISO/IEC 42001 Annex A.6 AI system life cycle and A.9 use of AI systems.
- Security and logging
- NIST AI RMF Measure and Manage. ISO/IEC 42001 Annex A.6, including A.6.2.6 operation and monitoring and A.6.2.8 event logs.
- Incidents and evidence
- NIST AI RMF Manage. ISO/IEC 42001 Annex A.8 information for interested parties, including A.8.3 and A.8.4, and clause 10.2.
What the score is for
- It shows which of your answers a buyer could challenge. Our page on AI security questionnaires explains how to answer from evidence.
- It is not a legal opinion. Whether a law applies to you, and in which role, is a question for your lawyer. The EU dates on this page are as of 20 November 2026.
- Where the answer is a document, you can write it. Where it is a product setting, an audit log or a guardrail, it has to be built. Read more in the questions AI agents get asked.
Questions before you start
Does a good score mean we are compliant or certified?
No. It is a self-assessment of 25 questions, not an audit. Nothing here is a certificate, and Mono does not certify anyone or give legal advice. ISO/IEC 42001 certificates come only from certification bodies accredited for that standard, and NIST has no certification at all. The check shows where your own evidence is thin.
Is anything I enter stored or sent?
No. Your answers stay in your browser's memory while the page is open. We don't store them or send them anywhere, and closing the tab clears them. Save as PDF uses your browser's print dialog, so the file stays on your device.
Which frameworks is the check based on?
The questions follow the shape of the NIST AI Risk Management Framework 1.0 (a revision is in progress) and the controls in ISO/IEC 42001:2023, and they cover what enterprise buyers ask in AI security questionnaires. It is not a formal mapping to either. Each area names its closest match so you can see where to read more.
Can the EU AI Act apply to an Indian SaaS company?
Yes, it can. The Act covers providers placing AI on the EU market wherever they are established, and third-country providers and deployers whose AI output is used in the Union. An Indian SaaS vendor selling an AI feature to EU customers is usually a provider, and the EU customer is the deployer. The AI Omnibus moved the high-risk deadlines to December 2027 and August 2028, but obligations remain. Ask your lawyer which tier applies to you.
What should we fix first?
Start with what a buyer asks for first: the inventory of AI features and vendors, the data flow map, vendor terms on training and retention, and an incident plan. Policies follow once you know what they have to cover. Mono prepares companies, builds the engineering and collects the evidence, and your lawyer and auditor do the rest.