Supplemental Terms for LLM AI Charts

Supplemental Terms for the "LLM AI Charts" Power BI Visual

Revised: August 9, 2026

These supplemental terms ("LLM AI Charts Terms") apply to your use of the "LLM AI Charts" Power BI custom visual ("the Visual", "LLM AI Charts") made available by CodeX Enterprises LLC ("CodeX", "we", "us", "our") (doing business as "BI-Champ" and "bizintelligencechampions.com" and "Biz Intelligence Champions"). They are IN ADDITION TO and form part of our general Terms of Use & Privacy Policy at https://bizintelligencechampions.com/terms. Where these supplemental terms address a topic differently than the general Terms, these supplemental terms control with respect to the Visual.

Your installation, use, or generation of charts through the Visual constitutes acceptance of these terms. If you do not agree, you must not use the Visual.

1. SCOPE

a. These terms cover the LLM AI Charts visual specifically. They cover: the in-Power BI client behavior; the server-side request handling, model selection, code generation, and quality-check pipeline; the freemium tier; usage credits and licensing; and persistence of any cached generated code in your Power BI report files (".pbix" / ".pbip"). Use of the broader bizintelligencechampions.com website, your Account, and any other Service is governed by the general Terms of Use.

b. Calling the engine from your own software is covered elsewhere. Where you reach our backend through our API, our Model Context Protocol (MCP) server, or an SDK — rather than through the Visual or another client we publish — the RegiaBI Programmatic Access Terms apply, and several sections of these terms are incorporated into them by reference. Programmatic access requires an active license; there is no freemium tier for it.

2. SERVICE MODES — HOSTED, BYO, FREEMIUM

a. Hosted mode (default). The Visual sends each generation request to our backend at bizintelligencechampions.com, which dispatches the request to a third-party large language model (LLM) provider. Routing depends on the model selected for the request: we may route through Microsoft Azure AI Foundry (which itself dispatches to upstream providers such as OpenAI, Anthropic, and DeepSeek), OR we may route directly to a provider's native API (for example, the Anthropic API, the OpenAI API, or a provider-specific Azure OpenAI deployment). Which routing path is used for which model is a backend implementation detail that may change at any time. The list of available models, their pricing, their relative cost, and their default selection may change at any time without notice, and any given model's availability may vary independently — subject to both our decisions and the upstream provider's availability and outages. We do not control or guarantee the availability, accuracy, latency, or behavior of the underlying LLM services and are not liable for interruptions, errors, or third-party policies.

b. Bring-Your-Own (BYO) mode. You may configure your Account to use your own API key against your own LLM provider account (for example, your own Anthropic, OpenAI, or Azure OpenAI account). In BYO mode, the request payload — including the constructed prompt, your dataset metadata, your sample data (paid only, see Section 4), and any usertext / commentary you have entered — flows from our backend directly to your chosen provider under your contract with that provider, not ours. We do not control or take responsibility for that provider's policies, billing, retention, abuse review, model behavior, or data handling. The LLM token cost in BYO mode is borne entirely by you with your provider. We do NOT charge usage credits for the provider's token cost itself; however, we DO charge a small flat per-request operating fee in usage credits for each BYO generation (configuration-driven and subject to change at any time without notice). Standard Account, service, and platform terms continue to apply.

c. Freemium tier. The Visual includes a freemium tier that allows new and casual users to evaluate it without an Account or purchase. The freemium tier is a privilege we extend at our discretion, not a contractual entitlement, and the following apply:

• Each Power BI visual instance receives an internally-generated freemium identifier. Usage credits and request counts are tracked against that identifier on our backend.
• Freemium has a per-instance spending cap, expressed in usage credits. We may change the cap, the unit of measurement, the formula for credit consumption, or the underlying model selected for freemium requests at any time without notice. A request that would exceed the remaining freemium budget will be refused; you will see a "freemium capacity exhausted" or similar message in the Visual.
• Freemium model selection is limited to a freemium-eligible subset of our hosted catalog. Bring-Your-Own (BYO) connections are not available in freemium mode, and our most capable / premium models — for example, Opus- and Fable-class Anthropic models — are excluded from freemium. If you do not select a model, we assign a freemium default. The freemium-eligible set, the default we assign, and the prompt configuration we use for freemium may change at any time without notice.
• Some Visual capabilities are reduced or disabled in freemium mode. Examples include but are not limited to: free-text commentary / user prompts ("USERTEXT") are stripped server-side before reaching the LLM and have no effect on generation; sample dataset rows are not transmitted to the LLM; some of the additional quality-assurance processing that paid tiers receive is reduced or disabled in freemium and should not be relied upon; user-history features such as version unlock and clone may be unavailable.
• Single-attempt generation — no automatic retries. On freemium, each generation request gets one shot. The automatic retry / correction that paid tiers receive on a failed generation does not run for freemium. A freemium request that does not pass our quality checks is shipped best-effort or refused rather than retried, so freemium output is more likely to contain an occasional rendering or quality defect than paid output. This is a deliberate quality limitation of the free tier, distinct from the credit cap and throttle above.
• We may disable freemium for any individual instance, IP address, or country at any time, including in response to suspected abuse, automation, or sharing of freemium identifiers across many reports.
• Freemium authoring availability is not guaranteed. Freemium authoring — generating new or revised chart versions through the LLM — may be throttled, paused, or made unavailable entirely at any time and for any duration we deem necessary, at our sole discretion (for capacity, cost, maintenance, abuse mitigation, or any other reason), with or without notice. This does NOT remove access to chart versions that have ALREADY been generated: once a chart version has been produced, it remains accessible and continues to render from its cached code in your saved report, independent of freemium authoring availability (see Section 5(c)).
• Freemium identifiers and any associated usage state may be reset, regenerated, or expired by us at any time. The freemium tier is not transferable.
• Freemium is intended for genuine evaluation and casual personal use. Use of freemium for production reporting, embedded distribution, or any commercial purpose is not permitted. For those uses, an organizational license is required (see the general Terms, Section 3).

d. Licensed (paid) tier. Use of the Visual under an organizational license consumes usage credits granted to or purchased by your Account. Cost per generation depends on the chosen model, prompt size, number of retries, and whether multi-pass features (validator, proofread) are engaged. We may change credit consumption rates, model availability, default model selection, prompt content, multi-pass thresholds, and other generation-pipeline behaviors at any time without notice. Credit balances reflect the configuration in effect at the time of each request and are not retroactively adjusted when configuration changes.

e. Feature matrix by service mode. The table below summarizes how features vary across the five service modes. Features that are the same across all modes (cross-filter, tooltips, cached re-rendering of saved reports, the in-browser code execution model, the tamper-detection signature, etc.) are omitted to keep the table readable; those are covered in the relevant sections of these terms and the general Terms of Use. The table is informational — specific behaviors may change as we tune the Visual (see Section 7); the section text above and below is the binding statement of these terms.

3. DATA SENT TO LLM PROVIDERS

For each generation request, the Visual transmits the following to our backend, and our backend (or your BYO provider, in BYO mode) transmits to the chosen LLM:

• Dataset shape metadata. Column names as they appear in your Power BI model, column data types, whether each column is a measure or dimension, the column's number/date format string, and aggregate statistics (counts, min, max, mean, median, standard deviation, quantiles, top category counts, and similar shape descriptors).
• Category values. Your actual text category values are obfuscated before transmission, apart from limited exceptions required so generated code can reference and order the chart's categories correctly. Where any value is transmitted, we do NOT promise it is protected against a determined reverse-engineering attempt; do not place secrets in category values.
• Sample dataset rows (higher privacy levels, paid licenses only). When you raise the privacy level to include a data sample, a small sample of rows is transmitted with non-numeric values obfuscated; numeric values pass through. We do NOT promise this obfuscation is irreversible; do not rely on it to protect secrets. Sample rows are NOT transmitted on freemium requests, and are not transmitted at all unless you opt in via the privacy setting.
• Viewport, settings, and request hints. Visual width and height in pixels; chart preferences (chart type hint, "favor traditional / non-traditional" preference, legend placement preference, derived-measure preference, "use similar / use different from last" preferences, font family choice); locale; and other generation-control hints derived from the Visual's settings pane.
• Your commentary / user text (paid licenses only). Any free-text commentary, per-field descriptions, or "describe any changes you want to make" text you have entered into the Visual's settings panel. This text is stripped server-side on freemium requests and has no effect on freemium generation.
• Diagnostic context for retries. When a request retries (for example after a code-quality rejection or a runtime exception), the previous attempt's error and a sketch of the violating code line may be sent on the retry prompt for the LLM to reason about its prior mistake.

We do NOT transmit the full unaggregated contents of your dataset to the LLM. We do NOT transmit your Account credentials, license key, or payment information to the LLM.

In hosted mode (Azure AI Foundry routing or our direct routing to a provider's native API), the underlying provider's terms govern how the transmitted content is handled. Such content is NOT used by us or by the underlying provider for model training. Prompts may be temporarily stored and reviewed by the underlying provider if flagged for abuse, in accordance with that provider's published terms.

In BYO mode, all content described above flows to your chosen provider under your contract with that provider, not ours. Data handling, retention, abuse-review policies, and training-eligibility in BYO mode are governed by your provider's terms.

4. DATA WE LOG ON OUR SERVERS

For every request to our backend (hosted mode and BYO mode alike, freemium and paid alike), we log on our own servers:

• Request metadata: timestamp, IP address, license key (if any), freemium identifier (if any), chosen model, request tokens, response tokens, credit cost, and a server-side correlation ID.
• The full constructed prompt as sent to the LLM, and the full response received from the LLM, including any quality-pass output.
• The LLM-generated code and a thumbnail (if available) of the rendered chart, when your Visual reports the post-render outcome.
• Runtime outcome from your client (success, exception type, stack trace), reported back to us when the Visual finishes rendering or fails. Used for diagnosis and quality measurement.
• Internal generation-pipeline telemetry and quality signals used to monitor and improve output quality.
• For BYO mode, we do NOT log the response payload returned from your provider beyond the metadata necessary for billing reconciliation and abuse detection.

We use these logs for diagnosis, billing reconciliation, abuse detection, and service improvement (including improving prompts, code-quality rules, and pipeline heuristics). We do not use these logs for training third-party models, and we do not sell them. We may share aggregated, non-identifying usage statistics in marketing materials or roadmap discussions.

5. CODE GENERATED BY THE VISUAL

a. Output is LLM-generated and may be wrong. Generated code (Python, JavaScript, JSON / Vega specifications, or HTML) and accompanying text are produced by large language models and may be inaccurate, incomplete, biased, off-by-one, hallucinated, or contain code that fails to execute or produces a misleading visualization. We make no representations or warranties as to the correctness, safety, or reliability of LLM-generated output. You agree to independently validate the resulting chart and any supporting analysis before relying on it for any business decision. This continues to apply even though we operate a multi-stage quality-check pipeline (validator pre-pass, code-quality regex checks, deterministic code patches, optional proofread pass, post-render visual stats): these stages reduce but do not eliminate the risk of incorrect or misleading output.

b. Code executes in your browser, not on our servers. The generated code is executed inside your Power BI session in your browser via sandboxed runtimes (e.g., pyodide for Python, an inline Function constructor for JavaScript, or direct DOM rendering for HTML / SVG). We do not execute LLM-generated code against your dataset on our servers. Risks introduced by the generated code at execution time (memory consumption, slowdowns, side effects within the visual's sandbox) are borne by your local environment.

c. Cached code is persisted in your saved report. When a generation succeeds, the resulting chart code, plus a small amount of associated metadata, is persisted into the Power BI report file (".pbix" / ".pbip") so that the Visual can re-render the same chart without making another LLM call. This persistence:

• does not depend on the report being open in Power BI Desktop — the cached code re-renders in any environment that loads the saved report, subject to license-status checks the Visual performs at render time;
• is integrity-checked. If the cached code in a saved report is edited outside the Visual, the Visual will refuse to render the modified code and will direct the user to the bizintelligencechampions.com "Unlock Version" feature to make supported edits. This integrity check is provided on a best-effort basis and should not be relied upon as a security control;
• may, for licenses that have expired or accounts that have changed, be gated by status checks the Visual performs at render time. Cached code from a freemium generation is not license-gated, but freemium and trial rendering still requires our backend to be reachable; cached code from a paid generation may require a valid license or an explicit unlock to continue rendering.
• Air-gapped / offline rendering. When our backend is temporarily unreachable, an explicitly UNLOCKED paid version is treated as self-contained and continues to render from its cached code — "air-gapped" use. All other cases — locked paid versions, trial versions, and freemium — require our backend to be reachable to confirm status before rendering; while it is unreachable, they show a "couldn't reach the licensing service" notice rather than rendering.
• Credential persistence and pinning. Except where a version has been pinned, a saved report may retain the credentials the Visual used to authenticate the generation — which can include your organizational license key and the associated account secret — so that the report can continue to render and generate. Pinning a version strips these credentials from the report file and locks the report to that specific cached version, producing a file that renders from its cached code without carrying your license key or secret. You are responsible for pinning any version before you distribute, share, or publish a report outside your organization; a report that has not been pinned may expose your credentials to anyone who obtains the file.

d. Code ownership. Subject to your compliance with these terms and the general Terms, you may use the generated code within your Power BI report for the purposes the Visual is designed for. You may not extract the generated code to use independently of the Visual, redistribute it as part of a competing product, or use it to train another model.

6. COST AND CREDITS

a. Credit consumption. Each successful generation consumes usage credits (paid) or freemium budget. For HOSTED requests, credit cost is computed from the chosen model's per-token rates and the actual input and output token counts (this token-based pricing replaced an earlier fixed-per-request model), plus any retry or quality-pass token consumption — so cost varies by model and request size. Credit cost is therefore NOT fixed or fully predictable per request: the model's response length, the number of retries, and whether quality passes engage are all determined during generation itself, so two similar — even identical — requests may consume different amounts of credits. Some cost drivers are within your control (which model you choose, how many fields you supply, how much commentary you enter); others are not. Treat published per-model rates as indicative of typical cost, not a fixed per-request price. For BYO requests, your provider bills you for the tokens directly and we instead charge a small FLAT per-request operating fee in credits. We publish current per-model rates on the Buy page and we may change those rates, the BYO per-request fee, or the credit-cost formula itself at any time without notice. Such changes apply to requests issued after the change takes effect; they do not retroactively adjust your credit balance or prior charges.

b. Retries and quality-check loops can consume more credits than expected. Our generation pipeline may issue additional LLM calls during automated quality-assurance and correction steps. Each of those calls consumes credits. We make commercially reasonable effort to cap runaway retry loops but do not guarantee any specific upper bound on credits per request.

c. Refunds. Consistent with the general Terms, we do not offer refunds for usage credits consumed by successful or failed generation requests. The freemium tier allows for evaluation of the Visual and its quality before any paid purchase.

d. Version unlock / clone / pin pricing. Creating a new editable version from your saved chart history is charged per new version, to your Account credit balance: unlocking a locked original costs 20 credits, while cloning an already-editable version costs 20 credits. Each new editable version is charged independently. Pinning a version for distribution costs 4,000 credits — pinning permanently removes the license requirement from deployed copies of that version (they render without any license, forever), which is why it is priced separately from, and higher than, unlocking. These charges are separate from the per-generation credit cost described in 6(a), and apply only when you create a new editable version (or pin one) through the bizintelligencechampions.com history page. The amounts are configuration-driven and may change at any time without notice. (Re-syncing an edit you make to a version you already unlocked is free.)

e. Freemium-to-licensed conversion. When a visual that was used in freemium mode is first upgraded to a licensed Account, the free freemium allowance is honored, but any freemium usage beyond that allowance is charged to your Account's credit balance as a one-time adjustment at the time of upgrade. This recovers our cloud cost for freemium generations that exceeded the free base before the visual was licensed. The allowance and the amount charged are configuration-driven and may change at any time without notice.

7. AVAILABILITY, RATE LIMITING, AND CHANGES

a. No service-level agreement. The Visual depends on third-party LLM providers, third-party cloud infrastructure, and our own backend. We offer no service-level agreement, uptime guarantee, latency guarantee, or response-quality guarantee. Service may be slow, partially degraded, or unavailable at any time without notice.

b. Rate limiting and throttling. We may throttle, queue, delay, or refuse requests at any time, from any IP address, account, or freemium identifier, for any reason (including but not limited to capacity management, abuse mitigation, or cost control). Decisions are at our sole discretion.

c. Configuration changes. We operate a database-driven configuration for many aspects of the Visual's behavior, including (but not limited to): which models are available, model pricing, model selection defaults, freemium spending caps, freemium model selection, prompt content and structure, code-quality rules, deterministic code patches, validator behavior, proofread thresholds, viewport-aware directives, chart-catalog membership, and runtime-weight biasing. We may change any of these at any time without notice, and changes take effect on the next request. Your continued use of the Visual after such changes constitutes acceptance.

d. Backward compatibility. A given saved Power BI report's cached chart will continue to render via its cached code as long as the data shape it expects remains compatible. We do not guarantee that REGENERATING that chart at a later date will produce the same code, the same chart type, the same derived measures, or the same visual appearance.

8. PROHIBITED USES SPECIFIC TO THE VISUAL

In addition to the prohibited uses in Section 2 of the general Terms, you may not:

• Use the Visual with data subject to the Health Insurance Portability and Accountability Act (HIPAA), Payment Card Industry Data Security Standard (PCI DSS), the GDPR special categories of personal data (Article 9), or any other regulatory framework that prohibits transmission of the relevant data to third-party LLM providers.
• Use the Visual with classified, export-controlled, or otherwise legally restricted data.
• Attempt to extract, infer, or reverse engineer the system prompt, validator prompt, proofread prompt, code-quality rules, or model selection logic for the purpose of building a competing service.
• Automate the freemium tier (e.g., scripted instance creation, freemium key cycling) to obtain free LLM access at scale.
• Use the Visual to generate code that targets, scrapes, attacks, or otherwise interacts with systems outside the Power BI render context the Visual provides.
• Use the Visual to produce content that violates applicable law or third-party rights.
• Share, transfer, or expose your BYO API key in a way that allows other users to make LLM calls on your provider account.

9. SUSPENSION AND TERMINATION

We may suspend or terminate your access to the Visual at any time, including (without limitation) for breach of these terms, suspected abuse of the freemium tier, automation in violation of Section 8, excessive cost generation, or risk to our service. We may also disable a specific saved-report instance's cached code from re-rendering when we have a reasonable basis to suspect tampering, sharing, or other prohibited use.

10. DISCLAIMERS, LIABILITY, AND INDEMNITY

All disclaimers of warranties, limitations of liability, and indemnity obligations in Sections 6, 7, 8, and 10 of the general Terms apply to your use of the Visual without modification. For clarity, and without limiting the general Terms:

• We are not liable for any business decision made in reliance on a chart, statistic, or analytical claim produced by the Visual.
• We are not liable for any cost incurred with a BYO provider arising from your use of the Visual, including cost from runaway retries, oversized prompts, or unexpected pricing changes by your provider.
• We are not liable for the actions of any third-party LLM provider, including but not limited to outages, billing errors, data-handling decisions, abuse-review decisions, or content-policy enforcement.
• We are not liable for any data the Visual transmits in accordance with Section 3, including data transmitted to a BYO provider, where you have configured the Visual to do so.
• We are not liable for any credit consumption, generation activity, account access, or other loss arising from your distribution, sharing, or publication of a report file that you did not pin. Pinning is provided as the mechanism to remove your license key and account secret from a report before it leaves your control; a report that has not been pinned may contain those credentials, and their exposure and any resulting use is your responsibility.

11. REVISIONS TO THESE TERMS

We may revise these LLM AI Charts Terms at any time by posting an updated version at https://bizintelligencechampions.com/terms-llmchart. Revisions are effective when posted. Where the change is material, we will use commercially reasonable effort to notify Accounts via the email on file. Your continued use of the Visual after revisions take effect constitutes acceptance.

12. CONFLICT WITH GENERAL TERMS

To the extent there is a conflict between these LLM AI Charts Terms and the general Terms of Use, these LLM AI Charts Terms control with respect to your use of the Visual. All other provisions of the general Terms continue to apply.

13. OPEN-SOURCE LIBRARY ATTRIBUTION

Open-source library attribution for the Visual (D3.js, Plotly.js, Vega / Vega-Lite, Pyodide and its Python scientific stack, and various browser-side utility libraries) is consolidated with the attribution for our other visuals in Section 14 of the general Terms of Use. Each listed library is distributed under a permissive open-source license (MIT, BSD-2-Clause, BSD-3-Clause, Apache License 2.0, ISC, Mozilla Public License 2.0, or Zlib) that permits commercial use.