NIST AI RMF · For AI governance leads at software firms
Your AI ships every month. Its risk record ships with it.
EA drafts your record against NIST’s AI Risk Management Framework for each AI system you sell, from your model inventory, product documents and evaluation reports. Where a test result is missing, it asks the engineer who holds it. Your AI Governance Lead approves each system’s record, and customers’ AI questionnaires are answered from it.
Try EA free. No card needed.
Organisations that run work on EA
The work
The models move faster than the paperwork about them.
Each release changes the answer
A new model behind the meeting summaries changes what they get wrong, and last quarter’s evaluation no longer describes it.
The knowledge sits with a few engineers
The evaluation sets, the red-team notes and the model versions live with the machine learning team, who are busy building the next release.
Customers ask in the framework’s shape
Enterprise buyers now send AI questions sorted into govern, map, measure and manage, and each wants them answered for the feature it is buying.
Kept system by system as each release ships, the record is ready before the buyer asks, and your machine learning team stays on the build.
What EA does
One AI system, through the four functions.
Shown on a meeting-summary feature that a software firm sells to enterprise customers.
01 / 04
- EA
Reads your AI policy, roles and approval routes, and lines each up against the framework’s governance outcomes, noting where no document says who is accountable. Knowledge Library
Your teamThe AI Governance Lead confirms who owns each system and each decision.
- EA
Drafts each system’s context from your product documents and model inventory, its intended use, users, data and the people its output reaches, with the source beside each line. EA Agent
Your teamThe Product Manager corrects the intended use where the product has moved on.
- EA
Sets each risk, including the Generative AI Profile’s, against the evaluation that measures it, and marks “Needs evidence” where none does. It sends the Machine Learning Engineering Lead one request for the missing results and follows up. EA Agent
Your teamThe Machine Learning Engineering Lead attaches the new evaluation for speaker attribution.
- EA
Drafts each risk’s treatment, monitoring and incident route, brings the system’s record to the AI Governance Lead, and answers the customer’s questionnaire from the approved record. Approvals Autofill
Your teamThe AI Governance Lead approves the record, and your team returns the customer’s questionnaire.
See your model inventory and evaluation reports drafted into one system’s record, with each missing test already asked for.
Book a demoNext release
A new model reopens only the lines it touches.
A model changes
Replace the model card or evaluation report on file, and EA redrafts the risk entries that cited the old one for their owner to check.
A feature launches
EA drafts the new system’s map from its product documents and carries over the policy, roles and approval route already approved under Govern, so only what is new goes to an owner.
A customer comes back
When a customer’s risk team questions an answer, EA drafts the reply from the same approved line and its evaluation, so the second answer matches the first.
- Risk
- Confabulation
- Evidence
- Evaluation report, model of 3 November
- Redrafted
- After the model change; 19 other entries left as they were
- Owner
- Machine Learning Engineering Lead
- Approved by
- AI Governance Lead Current
EA’s own security
Your evaluation results stay with the team that runs them.
Red-team notes and evaluation results show where your AI falls short today, so each sits in a workspace your AI Governance Lead controls. We ran our own ISO 27001 certification on EA: read how it ran on the ISO 27001 page.
- Hosted your way
EA’s cloud in a region you choose, hybrid options or on-premises. Data never used to train the models EA runs on.
- Access
Single sign-on, which you can enforce, and an access list for each workspace.
- Support access
EA staff see your data only in a time-limited window you open.
- Testing
An independent firm runs a penetration test on EA every year.
- Security contact
EA’s security team is at security@eaglobal.ai.
In the demo, see your red-team notes stay in a workspace of their own, with only the answer drawn from them going out to the customer.
Book a demoFAQ
Before the first system’s record goes to a customer.
A written account, for each AI system, of how you govern, map, measure and manage its risks: for a meeting-summary feature, who owns it, what it is for and whose meetings it summarises, how it was tested for confabulation, and what is done about it. The framework is voluntary, and your customers shape their AI questions around its four functions. EA drafts the record function by function from your own documents, adding the Generative AI Profile’s twelve risks for generative features.
Your customers’ AI and third-party risk teams, through their questionnaires and follow-ups; no body certifies against the framework. If a customer wants a certificate, ISO 42001 is the standard for that, and much of the same record carries over.
Each line of the record shows its source. If a customer asks how you measure bias in speaker attribution and no evaluation covers the current model, EA marks the line “Needs evidence” and asks the Machine Learning Engineering Lead, and nothing reaches the customer until your AI Governance Lead approves it.
Yes, whether that’s a model inventory, an AI policy or past customers’ AI questionnaires. Our Customer Success team helps with exports from other platforms. Plans differ by the number of autofills, the workspaces you need and the size of your knowledge library; see the pricing page.
NIST says version 1.0 is being revised as part of the White House AI Action Plan. As at 16 October 2026 no draft had been published, so EA drafts against version 1.0 and the Generative AI Profile of July 2024.
Your next model change, without rewriting the record around it.
One AI system’s record, to your AI Governance Lead.
AI governance, and the security work around it.
Related solutions
- ISO 42001 and the AI ActCertification for the same AI systems, and what the EU’s AI Act asks of you.Security and IT
- NIST CSFThe same publisher’s security framework, as profiles rated now and against a target.Security and IT
- Vendor security assessmentsCustomers’ wider security reviews, which now carry a section on AI.Security and IT
By industry
SoftwareAI features sold to enterprise buyers, each asked about in its own questionnaire.
Banking and building societiesBanks asking their suppliers how each AI feature is tested before it touches a customer.
Public sector and defenceSelection questionnaires, declarations and the carbon reduction plan, signed off by the right director.
Asset and investment managementInvestors’ DDQs, operational due diligence and the Principles for Responsible Investment report.
InsuranceBroker questionnaires, Lloyd’s confirmations and the own risk and solvency assessment.
Construction and infrastructureScheme renewals, prequalification and ISO evidence, including international engineering, procurement and construction projects.
Manufacturing and consumer goodsCustomers’ supplier questionnaires, Sedex and the conflict minerals template.
Energy and utilitiesThe CSRD statement, CDP, regulator returns and the network and information security self-assessment.
Healthcare and life sciencesPharma customers’ supplier questionnaires and the NHS data security toolkit.
Legal firmsPanel questionnaires, the professional indemnity renewal and Lexcel evidence, signed off by a partner.
Accounting firmsClients’ security questionnaires and the firm’s anti-money laundering return.
ConsultingClients’ vendor due diligence, framework tenders and the certificates behind them.

