Your own models

Your Own Models, the Same Checks

Embedded block-level LLMs (BYOM) with integrated Agentic RAG.

Syllabium is an enterprise learning management system (LMS) for regulated industries, made by Athantic AI LLC.

Bring your own model (BYOM) means that the AI inside a learning management system (LMS) can run on a model your organization brings, not only on the one the vendor chose. Syllabium is built for it: AI embedded in each block of a lesson, on your own model or ours, answering from your field’s own library. Whichever model answers, the request passes the same checks, stays inside the same limits and is logged the same way.

What does bring your own model mean in Syllabium?

In Syllabium, bring your own model means your organization’s AI requests can go to a model you bring, through your own AI account, instead of to ours. Ours are Claude models, through Amazon Bedrock and Anthropic's API.

Routes are set at three levels: the platform, each tenant (the owning company) and each organization inside it. A request goes by its organization’s route where one is set, else by its tenant’s, else by the platform’s. Where no route is set at any level, no call is made. Each model provider is held in the platform as data, not written into its code.

QuestionFor your own model and for ours
Which model answers?Your own model, through your own AI account, or ours: Claude models, through Amazon Bedrock and Anthropic's API.
What decides which one?The route set for your organization, else your tenant’s, else the platform’s.
What is checked first?The same twelve checks, in a fixed order, for every request.
What limits apply?Each agent’s limits on actions, tokens and spend, and a cap across all agents at each level.
What is recorded?Every attempt, with the check that decided it, and every model call, with its model, tokens and cost.

Which Models Does Syllabium Use?

Syllabium uses Claude models, through Amazon Bedrock and Anthropic's API. Where your organization brings its own model, its requests go to that model instead.

The name of a model says little about who controls it. What matters to a regulated organization is whose model answers, what it is sent, and who sets the rules it works under. Syllabium answers all three: the model can be yours, the platform assembles what it is sent, and your administrators set the rules.

How Is Your Own Model Governed?

Your own model is governed the same way as ours: every request passes the same twelve checks, in a fixed order, before any model is called. A request that fails a check is refused, the refusal names its check and is recorded, and the request never reaches a model.

ControlHow it works for every model
AuthorityFor each AI agent and each capability, an administrator sets one of four outcomes: Allowed, Human review, Admin review or Blocked. Unset means refused.
ScopeA grant can be limited to one library or one block.
Safety-critical contentOn content flagged safety-critical, an AI agent’s work waits for a person. No setting changes that.
DataThe platform assembles what the model is sent, and a person’s identity is never sent.
SpendEach agent has limits on actions and tokens an hour and on spend a day and a month, and every call has a cost cap, estimated before the call.
Emergency stopRead before every agent act, it can halt all AI, one agent, one kind of work or one type of content.
RecordEvery attempt is logged with the check that decided it, and every model call with its model, tokens and cost.

Governed AI shows the one route every AI action passes, and AI review before publish the review of AI work.

What Does Your Model Receive?

Your model receives only what the platform assembles for a defined action on named content, and never a person’s identity.

  • A rule in the database refuses identity data to every provider.
  • Authoring actions send content text only, and question checks send course content only.
  • A provider can be held to course and platform text alone, so that no learner’s writing reaches it.
  • A learner’s reflection goes to a model only under the learner’s enrollment consent, and without the learner’s name, email, user ID or cohort.
  • The action log’s chained rows hold no prompt, no answer and no person’s words.

Who sets the AI rules for your organization?

AI settings exist at three levels, the platform, each tenant and each organization, and a lower level can only be stricter than the one above it. The administrators of each tenant and each organization set their own level.

LevelWho it isIts AI settings
PlatformOperated by Athantic AI LLC.They limit every level below.
TenantThe owning company, licensed for a number of organizations.They can be stricter than the platform’s, never looser.
OrganizationWhere learning runs; it owns its content and works in its own field.They can be stricter than the tenant’s, never looser.

A cap across all agents applies at each level, beside each agent’s own limits. Loosening a safety-critical setting waits for a second, different administrator, and so does any change an agent proposes. A review of AI agents’ permissions is recorded, and is due every ninety days.

How Is Each Model Call Recorded?

Every model call is logged with its model, tokens and cost, in tamper-evident records kept seven years.

Tokens are the small pieces of text a model reads and writes, and model use is counted in them. Every attempt is logged as well, with the check that decided it, whether the act was allowed, refused or failed. Each entry carries a fingerprint of the one before it. Change one, and every later fingerprint stops matching. Once a day the newest fingerprints go to write-once storage outside the database, locked for seven years.

Administrators see AI spend on the console’s dashboard, a warning email goes out at 80 percent of a budget, and calls are refused at 100 percent. Evidence explains how the records are chained and checked.

Does bringing your own model change what AI may do?

No: the model answers, and the platform’s checks decide what may be asked of it, on which content, and whether a person must review the result.

In the authoring studio, each author sets the assistant to Off, Ask, Suggest or Act, and Suggest is the default. Authority can be limited to one library or one block, and above ordinary publishing an agent’s act always waits for a person.

Where Does Your Own Model Fit?

Your own model can answer wherever Syllabium calls a model: block-level LLMs, your model or ours, with agentic RAG grounded in your field.

Inside a lesson, it works block by block, with authority that can be granted down to a single block. Block-level LLMs explains how. When a person asks a question, agentic RAG answers from your field’s own library on the model your organization routes to, citing each claim to its paragraph and version. Agentic RAG explains retrieval and citations. Together they make one offering: enterprise learning with block-level LLMs, BYOM, and agentic RAG.

AI agents in training describes the agents that work inside Syllabium, and Platform the rest of the system.

Get Started

Enroll your organization online, see the pricing, or talk to us. Syllabium is made and operated by Athantic AI LLC.

Contact us