Legal

AI Use and Autonomy Disclosure

Effective date: 2026-05-07 · Last updated: 2026-05-06
In this document
1. What this disclosure covers2. What AXIS does autonomously3. AI providers used4. Accuracy expectations5. What AXIS does not do6. Your rights and controls7. EU AI Act compliance8. Bias and fairness9. Liability and warranties10. Changes11. Contact

This AI Use and Autonomy Disclosure explains how the AXIS Business

Operating System ("AXIS"), provided by Layer Zero Studios ("Layer

Zero", "we", "us"), uses artificial intelligence on your behalf, what

AXIS will and will not do autonomously, and the rights you have to

control AI use in your tenant.

This disclosure is written to satisfy the **EU AI Act (Regulation (EU)

2024/1689) Art. 50 transparency obligations for deployers** of

general-purpose AI systems, which become enforceable on 2 August 2026.

It supplements the Privacy Policy, the

Terms of Service, and the

Data Processing Agreement. **Lawyer review required

before publication.**

Primary-source references:

1. What this disclosure covers

AXIS is an AI-driven autonomous business operating system. It drafts,

suggests, and (where you authorise it) executes actions on your behalf —

sending emails, booking journal entries, generating plans, monitoring KPIs,

chasing invoices, and similar operational tasks. Because AI is at the

core of AXIS, you should know **what AXIS does on its own, where it can

be wrong, and how you stay in control.**

2. What AXIS does autonomously

2.1 Drafting

AXIS uses AI to draft content — emails, journal entries, campaign

copy, business plans, summaries, and similar artefacts. Drafted content

is shown to you (or to the operator-of-record on your tenant) before any

external action is taken, unless the relevant handler is on a higher

trust rung (see §3).

2.2 Acting

AXIS can execute actions on your behalf — for example, send an

email, create a Cal.com booking, post a Slack notification, write a row

to your finance ledger. Each action produces a receipt in

axis_action_receipts recording the provider, model used, status, proof

URL, and timestamp.

2.3 The trust ladder

Every handler in AXIS is rated on a three-rung trust ladder that

controls how autonomous it is. Each new tenant starts at L1 by default;

rungs are earned over time through approved-action history (TrustScore).

| Rung | Behaviour |

|---|---|

| L1 — Suggest only | AXIS proposes an action and waits for your explicit approval before doing anything externally. This is the default for every new tenant on every handler. |

| L2 — Act with receipts | AXIS acts and records a receipt; you are notified after the fact and may review or revert. Used only for handlers you have promoted explicitly. |

| L3 — Autonomous + human review | AXIS acts unsupervised; receipts are retained for your scheduled review. Used only for low-risk, high-volume handlers (e.g. invoice-chase reminders) and only after the handler has accumulated a sufficient TrustScore. |

You can withdraw autonomy on any handler at any time — see §6.

3. AI providers used

AXIS routes AI workloads to the providers below. The full sub-processor

list is at /legal/subprocessors.html;

the privacy mechanics are described in Privacy Policy §4

and DPA §6.

| Provider | Use | Personal Data sent? |

|---|---|---|

| Anthropic (Claude) | Primary LLM for drafting, planning, reasoning | Yes, scoped to handler need; covered by paid-API no-train terms |

| Google (Gemini) | Research handlers, intake question generation | Yes, scoped to handler need; covered by paid-API no-train terms |

| DeepSeek | Cost-optimised inference for non-sensitive tasks | No. Prompts are sanitised at the platform layer before egress |

We are inference-only customers of all three providers. Under our

contracts with them, none of your data is used to train their models.

If any of those guarantees breaks, we will notify you within 72 hours per

POPIA s22 and GDPR Art. 33.

4. Accuracy expectations

AI is powerful but fallible. AXIS adds platform-level safeguards on top

of the provider models:

(e.g. a journal entry, a draft invoice), AXIS runs schema checks and

cross-handler sanity checks before presenting it.

it surfaces the uncertainty and asks you, rather than guessing. This

is a design choice, not a guarantee.

axis_action_receipts row. Nothing is silent.

Despite these safeguards, AI output can still be wrong. **Common failure

modes you should plan for:**

fabricated invoice number, a wrong contact title).

current events, AXIS routes to research handlers, but those have their

own latency and coverage gaps.

may carry assumptions about industry, geography, or demography.

contain instructions intended to manipulate the model.

**You are responsible for reviewing material AI-generated content before

it is transmitted to a third party** — for example, before sending an

AXIS-drafted email to a customer of yours, or before filing an

AXIS-prepared journal entry with your bookkeeper. The L1 default is

designed to keep you in that loop by default.

5. What AXIS does not do

The following are out of scope. AXIS will not do these things, even when

asked or configured to:

medical, tax, or financial advice. Where a question requires

professional judgement, AXIS will surface the boundary and (when

available) route to the Tier-3+ Professional Marketplace of vetted

human professionals. Until that marketplace is live, AXIS will tell

you the question is out of scope.

confirmation.** Tax filings, statutory returns, employment decisions,

contractual commitments to a third party — all require **explicit

operator confirmation** at the L1 review step before AXIS acts.

of every AI provider listed in §3. Your data is not in their training

pipelines under our contracts.

drafts a message that AXIS will send on your behalf, AXIS may include

an "AI-drafted" footer disclosure when configured (see §8.1). You may

not configure AXIS to actively misrepresent AI authorship to your

counterparties.

This list mirrors the platform commitments in our internal AXIS_BLUEPRINT

§3.5 and is testable by reading any handler's source code.

6. Your rights and controls

6.1 Right to know

Every action AXIS takes on your behalf records the AI provider and

model used. You can query the receipt at any time via

Dashboard → Receipts or via the AXIS API.

6.2 Right to disable a provider

You can disable a specific AI provider for your tenant. For example, if

you do not want any handler to call DeepSeek, set

Dashboard → AI → Providers → DeepSeek = off. Affected handlers will

re-route to an alternative provider (typically Claude) at no extra cost.

6.3 Right to disable autonomy

You can downgrade any handler to L1 (suggest only) at any time. You

can also activate the freeze-all-autonomy kill-switch

(Dashboard → AI → Freeze) which forces every handler back to L1

immediately and pauses any in-flight action that has not yet executed.

6.4 Right to human review

Under GDPR Art. 22 and POPIA s71, you (or any data subject whose

data AXIS has processed) have the right to challenge an automated

decision that has legal or similarly significant effects, and to request

human review. Send the request to

privacy@layerzerostudios.com. We

respond within 30 calendar days with the action receipt, the AI

provider and model used, the inputs (with PII redacted as appropriate),

and the human-review outcome.

6.5 Right to a copy of receipts

The receipt log is exportable from

Dashboard → Account → Privacy → Export my data in JSON format,

including every AI-touched action.

7. EU AI Act compliance

7.1 Deployer status

Layer Zero is a deployer of general-purpose AI systems under the

EU AI Act Art. 3(4). We integrate third-party AI models (Claude,

Gemini, DeepSeek) into AXIS rather than training our own foundation

models. The Art. 50 transparency obligations therefore apply to

us as deployer.

7.2 Disclosure of AI interaction

Where AXIS sends a message on your behalf to a human end-user, AXIS may

add an "AI-drafted" footer disclosure to satisfy the Art. 50(1) and

Art. 50(4) requirements. This is configurable per handler. If you switch

the disclosure off, you accept the corresponding compliance

responsibility under the AI Act and any applicable national rules.

7.3 Synthetic content labelling

Where AXIS generates synthetic text, audio, image, or video content for

your tenant that is intended for public dissemination, that content is

labelled as AI-generated per Art. 50(2), unless you have switched the

labelling off and accepted responsibility (see §7.2).

7.4 Enforcement timeline

The Art. 50 obligations enforce on 2 August 2026. Layer Zero's

Art. 50 conformance is reviewed quarterly; the next review is logged in

this disclosure's revision date.

8. Bias and fairness

We monitor AI outputs for bias at the handler level — primarily through

manual review of receipts and through customer feedback. **You can flag

biased or unfair output via the in-dashboard feedback button on any

receipt. We triage flagged output within 5 business days** and, where

a pattern is identified, adjust the handler prompts or the routing.

We do not represent that AXIS is bias-free. AI bias is an active area of

research; we treat it as an ongoing engineering obligation rather than a

one-time guarantee.

9. Liability and warranties

AI-generated output is provided "as is". We make no warranty of

fitness for any specific decision. The limitation-of-liability framework

in the Terms of Service §11 governs.

Nothing in this disclosure waives:

review of solely-automated decisions

10. Changes

We will update this disclosure when a material change to AXIS's AI

behaviour is shipped — for example, when a new AI provider is added,

when a handler is promoted to a higher trust rung by default, or when

the EU AI Act guidance shifts in a way that changes our posture.

Material changes are emailed to active customers at least 14 days

before they take effect.

11. Contact

© Layer Zero Studios
legal@layerzerostudios.com layerzerostudios.com