ISO/IEC 42001 is the international standard for an AI management system, and Syllabium is built to it. Syllabium is an enterprise learning management system (LMS) for regulated industries, made by Athantic AI LLC. This page explains the standard in plain words, and what Syllabium is built to under it.
ISO/IEC 42001 is the international standard for an AI management system: how an organization governs the AI it builds and runs. ISO and IEC, the two international standards bodies behind it, first published it in 2023.
A management system is the set of policies, roles and routines through which an organization runs something and keeps improving it. ISO/IEC 27001 does this for information security, and ISO/IEC 42001 does it for AI. The standard asks an organization to set an AI policy, assess the risks and the impact of each use of AI, put controls on how its AI is built and used, audit itself, and have its management review the results.
An independent certification body audits an organization against the standard and, if it passes, issues a certificate. The certificate belongs to an organization, not to a product: for Syllabium, that organization is Athantic AI LLC, which makes and operates it.
Built to is a statement about design: it names a standard the platform is designed and built to. Certification to ISO/IEC 42001 is planned, and this page claims no certificate.
Toward the standard, Syllabium has a rule of one impact assessment for each use of AI, graded authority for every AI agent, an emergency stop, an action log in tamper-evident records, and a recorded review of each agent’s permissions. Athantic AI LLC’s AI management system itself, with its AI policy, internal audit and management review, is planned, and so are a register of AI suppliers and an assessment of the model suppliers.
An AI impact assessment is a written study of one use of AI: what it does, whom it can affect, what can go wrong, and which controls answer each risk. Syllabium’s rule is one impact assessment for each use of AI, and three were approved by November 8, 2026, each as a written document.
The table lists the uses of AI in Syllabium and how each one works.
| Use of AI | How it works |
|---|---|
| The authoring assistant | The assistant sits beside the editors in the authoring studio, in one of four modes each author chooses: Off, Ask, Suggest or Act. Suggest is the default. |
| The engagement check | Each day, an agent reads three counts for each learner (overdue lessons, missed reflections and days unseen) and raises a suggested intervention in fixed words. No model is called, no learner’s writing is read, and every raise waits for a staff member. |
| A draft note on a reflection | The platform drafts a note for a supervisor on a learner’s written reflection, and the learner never sees it. It is built and not yet in operation: it stays off until the enrollment consent names it and a person turns it on. |
| The question check | An AI verifier, and then a separate AI critic, check each assessment question against the slide text it cites. It is built and not yet in operation, and a rejected safety-critical question goes to a reviewer. |
| Block-level LLMs and agentic RAG | AI embedded in each block of a lesson, on your own model or ours, answering from your field’s own library. |
AI settings exist at three levels, the platform, the tenant and the organization, and a lower level can only be stricter than the one above it. The administrator of each level sets that level’s controls.
| Control | What it does |
|---|---|
| Graded authority | For each AI agent and each capability, an administrator sets Allowed, Human review, Admin review or Blocked. Unset means refused. |
| Levels of authority | Authority runs from viewing, through drafting and ordinary publishing, to safety-critical publishing, deleting and full authority. Above ordinary publishing, an agent’s act always waits for a person. |
| Scope | A grant can be limited to one library or to one block. |
| Limits and budgets | Each agent has limits on actions and tokens (the units in which a model’s work is counted) an hour, and on spend a day and a month. Every call carries a cost cap, estimated before it runs, and a cap covers all agents at each level. A warning email goes out at 80 percent of a budget, and calls are refused at 100 percent. |
| Emergency stop | Every AI action reads the emergency stop first. A stop covers everything, one agent, one kind of work or one type of content; it is set in two steps, with STOP typed again, and a red banner shows on every AI page while it holds. |
| A second administrator | Loosening a safety-critical setting, any change an agent proposes, and a loosened rule each wait for a different administrator. |
| Agent identity | Each AI role acts through its own account, with no password and no sign-in, and no person can hold an AI role. Staff see “(AI)” and the role beside an agent’s work. |
| A person in charge | A person activates or retires an AI role, with a recorded reason. An agent never approves, and the assistant does not appear on the pages that govern it. |
| Access review | A review of AI agents’ permissions is recorded, and is due every ninety days. |
| Personal data | Identity data is never sent to a model: a rule in the database refuses it for every provider. |
The action log records every AI attempt with the check that decided it, and every model call with its model, tokens and cost. Its entries are tamper-evident and kept seven years.
Every AI action passes twelve checks in a fixed order before it runs: the emergency stop, the switches, tenancy, the role’s state, its permission, the safety-critical rule, any override, the level of authority asked, the library’s scope, limits, the class of data, and the decision. A refusal names the check that refused it, is recorded, and never reaches a model. The record is written whether the action was allowed, refused or failed, and its chained rows hold no prompt, no answer and no person’s words.
The records behind the log are described on Evidence.
A person flags content as safety-critical, and on that content no setting gives an AI agent a plain allow: its work waits for a person. The rule is enforced in the code and in the database.
A change to content flagged safety-critical stays hidden from learners until a second person approves it, and the person who submitted it cannot approve it. An administrator can switch off the second person for people’s own changes, which itself takes a second administrator, but never for AI. In the authoring studio, the assistant does not publish content. The route from an AI draft to published content is on AI review before publish.
What ISO/IEC 42001 asks for, in plain words, maps to what Syllabium has as follows.
| ISO/IEC 42001 asks for | What Syllabium has |
|---|---|
| An AI policy and a management system, with internal audit and management review | Athantic AI LLC’s AI management system is planned. |
| An assessment of each use of AI | One impact assessment per use of AI; three were approved by November 8, 2026. |
| Controls on what AI may do | Graded authority for each agent and capability, twelve checks before every AI action, limits and budgets, and an emergency stop. |
| Human oversight | Safety-critical AI work waits for a person, and loosening a safety-critical setting waits for a second administrator. |
| Records of how AI operated | The action log, in tamper-evident records kept seven years. |
| Review of what AI may do | A recorded review of AI agents’ permissions, due every ninety days. |
| Rules kept on record | The platform’s rules sit in one register, and roles and capabilities are data, changed without a release. |
| Oversight of AI suppliers | A register of AI suppliers and an assessment of the model suppliers are planned. |
| Certification | Certification by an independent body is planned, and no certificate is claimed. |
No. This page describes how Syllabium governs the AI inside it. An organization that teaches the standard to its own people can author that course in Syllabium Studio, the authoring studio described on Courseware.
ISO/IEC 42001 is one of ten standards Syllabium is built to; among the others are ISO/IEC 27001 for information security, SOC 2 Type II for a service provider’s security controls, and NIST CSF 2.0 for managing security risk. Each is explained on Standards, and how the AI inside Syllabium works under authority is on Governed AI.
Enroll your organization online, see the pricing, or talk to us. Syllabium is made and operated by Athantic AI LLC.
Contact us