Glossary
EU AI Act Terms + Audact Patent Families.
The vocabulary a compliance lead actually uses. Each term linked to the page where it lives in practice.
A
- AI Act Omnibus
- Working informal name for the consolidated EU AI Act package and its implementing acts. Used in EU Commission communications during the 2025–2026 implementation window.
- Article 4
- EU AI Act provision on AI literacy. Requires deployers to ensure staff interacting with AI systems have appropriate AI literacy.
- Article 27
- EU AI Act provision requiring a Fundamental Rights Impact Assessment (FRIA) for high-risk AI deployment in scoped sectors.
- Article 50(1)
- EU AI Act provision: deployers of AI systems interacting with natural persons must inform them they are interacting with AI.
- Article 50(2)
- EU AI Act provision: providers of generative AI systems must mark outputs as artificially generated in a machine-readable manner.
- Article 57
- EU AI Act provision establishing regulatory sandboxes at member-state level for AI innovation under supervisory oversight.
- Article 86
- EU AI Act provision on the right to explanation of individual decision-making for affected persons.
- AudactRules
- Open executable specification for compliance rules, to be published by Audact under CC-BY-SA (v0.1 in development). Designed to let runtime engines interpret rules consistently across vendors.
C
- C2PA
- Coalition for Content Provenance and Authenticity. Open standard for cryptographically signed media provenance, foundational for Article 50(2) watermarking.
- Cryptographic Evidence Chain
- Hash-linked sequence of compliance events (disclosure, tool-call, policy-version) anchored in a Merkle tree and RFC 3161 timestamp. Audact's core evidence architecture.
D
E
- EUDI Wallet
- EU Digital Identity Wallet. Member-state-issued digital identity primitive enabling verifiable credential exchange across the EU.
- Evidence Vault
- Audact API (launching Q3 2026) designed to expose the cryptographic evidence chain as a queryable, exportable, tamper-evident artefact store.
F
- FRIA
- Fundamental Rights Impact Assessment. Required under Article 27 of the EU AI Act for high-risk deployments in scoped sectors.
G
- GDPR Art.28 DPA
- Data Processing Agreement under Article 28 of the General Data Protection Regulation, governing the relationship between data controller and processor.
I
- ISO 42001
- ISO/IEC 42001:2023 — international management-system standard for Artificial Intelligence. Audact targets certification with NEN/ISO 42001 alignment Q3 2026.
M
- MCP
- Model Context Protocol. Open standard by Anthropic for connecting LLMs to tools and data sources. Audact targets a native MCP server Q3 2026.
- Merkle Tree
- Cryptographic data structure where each non-leaf node is the hash of its children. Enables tamper-evident batch attestation of large event streams.
- MiCA
- Markets in Crypto-Assets Regulation. EU regulation on crypto-asset issuance and service provision. In force since December 2024.
N
- NL Telecomwet
- Dutch Telecommunications Act. Updated provisions on AI-assisted outbound calling take effect 1 July 2026.
P
- Per-Jurisdiction Policy Engine
- Runtime mechanism that selects the correct rule set (disclosure copy, recording consent, outbound-eligibility logic) based on caller and callee jurisdiction.
- PII Crypto-Shredding
- Technique for rendering personally identifiable information unrecoverable by destroying the encryption key, while keeping the surrounding ciphertext intact for audit.
R
- RFC 3161 Timestamp
- IETF standard for trusted timestamping. Audact anchors its evidence chain to RFC 3161 timestamps from a third-party time-stamping authority.
W
- W3C Verifiable Credentials
- Open standard for cryptographically verifiable digital credentials. Used in compliance attestations and identity flows.
Audact patent families
Fifteen of the 44 patent-pending families (UK IPO) that map most directly to the compliance vocabulary. Each links to the relevant cluster on /patents.
- Inline Proxy
- Inline compliance proxy that mediates AI conversations and produces an evidence record per interaction.
- Compliance Receipt
- Per-interaction signed receipt that captures disclosure, policy version, model identity and tool-calls.
- Multi-Candidate
- Mechanism for evaluating multiple candidate model outputs against policy before selection.
- Incident Reconstruction
- Cryptographically verifiable reconstruction of a past AI interaction from anchored evidence.
- Brand Governance
- Real-time enforcement of approved messaging and brand-safety rules against AI output.
- Commercial Transaction
- Pre-dispatch validation of AI-initiated commercial transactions against pricing and contract rules.
- Article 50 Attestation
- Mechanism for producing per-call Article 50 disclosure attestations across channels.
- Cross-Jurisdiction Consent
- Engine for handling consent and recording rules across jurisdictions in a single conversation.
- Scope-of-Practice
- Validation that AI agent output stays within the regulated scope of practice for licensed professions.
- Outbound AI Call
- Eligibility, disclosure and evidence requirements for AI-initiated outbound voice and chat.
- Vulnerable Person Detection
- Runtime detection of indicators of vulnerability with policy-gated escalation paths.
- Disclosure Sequencing
- Ordering and conditional logic for layered Article 50 disclosures in multi-turn conversations.
- Channel-Agnostic Attestation
- Attestation mechanism that produces the same evidence semantics across voice, chat and outbound channels.
- Outbound Eligibility Engine
- Pre-dispatch eligibility check combining jurisdiction, consent, do-not-contact and Article 22 logic.
- FRIA Generation
- Auto-generation of Fundamental Rights Impact Assessments from live deployment profile.
See the patents in context
The full ten-cluster map and the PCT roadmap live on /patents.