compliance · the UAE AI Charter
Wytness and the UAE AI Charter
Principle-by-principle evidence for the AI agents you develop or run in the UAE.
The UAE Charter for the Development and Use of Artificial Intelligence (UAE AI Office, July 2024) sets twelve principles for anyone developing or using AI in the UAE. It is a guiding framework with no certification. Where your AI is agents, Wytness evidences the 6 principles that rest on a record of what the agents did and how they were governed: safety monitoring, bias and impact assessments, data privacy, transparency of actions, human oversight, and governance and accountability. The other 6 are listed below with the reason. This is a first mapping; the pack evidences your governance record, not the models themselves.
Sources: The UAE Charter for the Development and Use of Artificial Intelligence (July 2024, PDF) · UAE Government portal page
the artefact
What's in the pack
A UAE AI Charter pack contains the agent and tool inventories, evidence and a measured metric for each mapped principle, signed event samples, the LLM provider and model mix, and the list of principles it does not evidence, with the reason for each.
Every claim in the pack is traceable: each control cites the events behind it, and each event carries the signature your own key produced. The pack is sealed, so an auditor can confirm it reached them unaltered.
A real sealed pack is published — nothing here is a mock-up. Download the specimen and check it yourself on the verify page, in your browser, without uploading it. It is a SOC 2 (Trust Services Criteria) pack — the seal and the per-event signatures work identically whichever framework you generate.
// What the pack asserts about itself:
// control_mapping[] — 6 criteria, each carrying
// its evidence and a metric measured over your period:
…// + 2 more, in the table below
// Every control cites events an auditor can re-verify:
the mapping
What we map to what
| Principle | Name | What Wytness provides |
|---|---|---|
| Principle 2 | Safety | 7 anomaly-detection rule families run against the live event stream (first-time tool use, capability drift, frequency spikes, error spikes, restricted-data access, off-hours activity, hash-chain breaks) with severity-classified alerts and a resolution state per alert. Each register row carries a risk classification (minimal / limited / high / prohibited, with a justification required for high and prohibited) and moves through the lifecycle Discovered → Intake → Development → Validation → Approval → Deployment → Retirement, so a system modified or withdrawn because of risk leaves a dated record of who moved it. Safety testing of the model is the customer's; the kill action records the request and does not stop the agent. |
| Principle 3 | Algorithmic Bias | The register's AI risk assessment includes a bias and fairness section, and the responsible AI impact assessment covers affected people and potential harms; each assessment is answered by the customer against a register row, scored on submit, and its gaps become remediations with an owner and a due date, with each submission and each remediation's creation, status change and deletion written to the audit log. Wytness records that the assessment was made and what it found; it does not test models or data for bias. |
| Principle 4 | Data Privacy | Where the customer enables PII keys, personal data in payloads is replaced with HMAC pseudonyms in the customer's runtime before egress and the token map is encrypted to the customer's X25519 public key, so Wytness stores pseudonyms it cannot reverse. Every event records inputs_source and inputs_classification, the register tags each agent with the data classes it accesses, an event classified restricted raises a restricted_data anomaly, and subject search and subject erasure let the customer act on a data subject's request. The privacy impact assessment and the lawful basis remain the customer's. |
| Principle 5 | Transparency | Every tool call, resource access and inter-agent message is logged with its inputs, outputs, tool parameters and the agent's reasoning_summary where the runtime supplies one, hash-chained (tamper-evident) and signed per event with the customer's Ed25519 key; sessions group the actions of one run so the steps behind an outcome can be read in order. Evidence Packs are paired with a sealed JSON an auditor verifies offline without a Wytness account. Explaining a model's internal decision process to the public is the customer's. |
| Principle 6 | Human Oversight | Every event records human_operator_id, and each register row carries an assigned operator; moving a row into each lifecycle stage requires the role the organisation sets for that stage, approval needs a second person (four-eyes) where the organisation has two or more admins, and each approval change is logged with its from / to states and the written reason the app asks for. Wytness records the oversight that took place; it does not gate agent actions at runtime. |
| Principle 7 | Governance and Accountability | The Agent Registry inventories every agent seen in the event stream (added as Discovered so none runs unrecorded) with its operator, intended use, risk classification and lifecycle stage; four-tier role-based access control (viewer / member / admin / owner) is enforced server-side on every route; administrative actions on the organisation, including register, approval, assessment and remediation changes, are written to the customer's Audit Log, which exports as CSV and can be forwarded to the customer's own SIEM. Evidence Packs document the period for auditors. |
the boundary
What Wytness does not cover
| Principle | Name | Why the pack does not evidence it |
|---|---|---|
| Principle 1 | Strengthening Human-Machine Ties | A statement of national intent about human well-being; no record of agent activity evidences it. |
| Principle 8 | Technological Excellence | A national innovation and competitiveness aim, outside what a governance record can show. |
| Principle 9 | Human Commitment | A values statement about placing human values at the heart of innovation; a customer can record its own position in a responsible AI impact assessment, but the commitment itself is not evidenced. |
| Principle 10 | Peaceful Coexistence with AI | A societal aim about human security and fundamental rights across communities, outside the product. |
| Principle 11 | Promoting AI Awareness for an Inclusive Future | Public awareness and equitable access to AI are aims for society; Wytness has no mechanism for them. |
| Principle 12 | Commitment to Treaties and Applicable Laws | Legal compliance is the customer's determination; the record supports it, but this pack does not evidence compliance with UAE law or international treaties. |
- —The charter certifies nothing, and neither does Wytness. An Evidence Pack is a principle mapping with evidence, not a statement of conformity.
- —Wytness does not test models for bias, safety or accuracy. The bias and impact assessments you record in the Registry are your own answers, tracked there and not attached to the pack; we do not grade the model.
- —Compliance with UAE law, including personal data protection law, is your determination. This first mapping covers the charter's principles only.