Contents
Standing caveat
This map lists obligations where TELOS records and measurements may be relevant. TELOS AI Labs Inc. wrote it as a self-assessment. It is not an independent certification and it is not a legal opinion. A technical artifact can support a compliance process and still not satisfy the governing obligation. Legal sufficiency depends on the regulated role, system classification, jurisdiction, deployment facts, record completeness, retention, and review by qualified counsel, auditors, and regulators.
The map is current as of August 11, 2026. Laws, implementation dates, guidance and regulators' positions can change.
Capability-to-evidence map
Constitutive fidelity is the measured correspondence of submitted work to a human-declared specification, using the versioned Purpose Anchor in force. A measurement like that may contribute evidence to a compliance process.
Scroll horizontally to read all columns.
| TELOS capability | Potential evidence contribution |
|---|---|
| Versioned Manifest | Declared purpose, scope, tools, sequences, boundaries, and approval requirements |
| Per-action measurement | A contemporaneous comparison between submitted work and a fixed external reference |
| Drift and stability telemetry | Post-deployment monitoring and trend evidence |
| Graduated surfacing | Evidence that TELOS flagged uncertainty or a departure for human attention |
| Public hash-linked projections | Payload and published order integrity |
| Published Record of Trust | Signed commissioning record with hash links, checkable offline |
| Retention setting | Declared intended retention period |
European Union AI Act
Primary source: Regulation (EU) 2024/1689.
Article 72: post-market monitoring
Providers of high-risk AI systems must establish and document post-market monitoring that actively and systematically collects, documents, and analyzes relevant performance data over the system’s lifetime. TELOS per-action scores, drift indicators, stability measures, and event logs may be relevant evidence because they observe behavior after deployment.
That contribution depends on the deployment. It needs to establish coverage, metric validity, record completeness, issue handling, retention and a documented process that evaluates continued compliance.
Articles 11 and 12: documentation and record keeping
The Manifest, configuration record, signed receipts, and telemetry can contribute to technical documentation and logging.
Article 14: human oversight
An escalate classification marks a departure for human review. To show that oversight happened, a deployment also needs the adapter event record, recipient, response, timing and resolution.
Article 15: accuracy, robustness, and cybersecurity
The retained agentic-benchmark results in the Validation Report are self-scored development evidence. They may contribute to an evidence package.
Article 19: automatically generated logs
Article 19 sets retention requirements for certain automatically generated logs under a high-risk provider’s control. TELOS exposes a retention setting. Deployers have to implement the applicable retention period, legal holds, access controls, and deletion behavior outside or around the current component.
Article 50: transparency
Article 50 includes transparency duties for specified AI interactions and generated or manipulated content. TELOS can preserve an account of what a governed agent submitted and how TELOS scored it.
Article 6, Annex III and Article 99
Classification and penalties depend on the system and regulated role. Annex III includes specified uses such as biometrics, critical infrastructure, education, employment, essential services, law enforcement, migration and border control, and administration of justice and democratic processes. If a TELOS component is embedded in a regulated system, assess it inside the actual system boundary.
EU AI Omnibus
The European Commission proposed changes to implementation timing in November 2025. Its AI Act page states that the AI Omnibus proposal was adopted on 19 November 2025, political agreement was reached on 7 May 2026, and the resulting regulation entered into force on 27 July 2026. Under that revised timeline, the high-risk rules in Annex III will apply from 2 December 2027 and those in Annex I from 2 August 2028.
Primary source: European Commission, AI Act regulatory framework.
California SB 53
Primary source: California SB 53 bill record.
California’s Transparency in Frontier Artificial Intelligence Act took effect January 1, 2026. It applies to covered frontier developers that meet statutory conditions, and it has provisions on safety frameworks, transparency reports, incident reporting and whistleblowers.
TELOS is aimed mainly at deployers, so this mapping is directional. Runtime evidence may help a covered organization show that it monitored declared limitations and evaluated adversarial behavior. Two questions need analysis of the actual entity, model, role and incident. The first is whether TELOS sits inside a covered developer’s safety framework, and the second is what evidence the developer must publish or report.
Colorado SB 26-189
Primary source: Colorado SB 26-189, enacted May 14, 2026.
Colorado SB 26-189 repealed and reenacted SB 24-205 with new requirements for automated decision-making technology used to materially influence consequential decisions. Principal requirements begin January 1, 2027.
The enacted summary identifies, among other things:
- technical documentation from covered developers describing intended uses, training-data categories, known limitations, appropriate use, and human review;
- notice obligations for covered deployers;
- plain-language descriptions following specified adverse outcomes;
- consumer rights concerning data, correction, meaningful human review, and reconsideration;
- at least three years of compliance records for developers and deployers.
TELOS records may support documentation of intended purpose, measured departures, configuration, and human review.
FDA Quality Management System Regulation
Primary sources: FDA QMSR overview and 21 CFR Part 820.
The Quality Management System Regulation became effective February 2, 2026. It amended 21 CFR Part 820 and incorporates ISO 13485:2016 by reference.
For a regulated finished-device manufacturer, TELOS telemetry may contribute to evidence about validation and monitoring of production or service processes, nonconformity handling, corrective action, and records. Applicability depends on the device, the use of AI within the quality system or product lifecycle, and the manufacturer’s validated procedures.
A QMS record must establish approved procedures, validation acceptance criteria, change control, review, investigation, corrective action, and record control. FDA record retention depends on the record type and the governing requirement.
ISO 9001:2015 and ISO 13485:2016
Primary sources: ISO 9001 and ISO 13485.
TELOS borrows two analogies, the Plan-Do-Check-Act cycle and statistical process control:
Scroll horizontally to read all columns.
| Quality phase | TELOS mapping |
|---|---|
| Plan | Approve a Manifest and measurement configuration |
| Do | Run the agent through an integrated measurement path |
| Check | Measure per-action fidelity, drift, stability, and departures |
| Act | Surface, investigate, resolve, recalibrate, or change the specification under control |
TELOS may be relevant to monitoring and measurement, the ability of processes to achieve planned results, nonconformity and corrective action. To conform to either standard you need an audited management system.
NAIC Model Bulletin on AI used by insurers
Primary source: NAIC model bulletin announcement and materials.
The bulletin expects an insurer’s AI program to address governance, risk management, documentation, and accountability. TELOS can contribute declared purpose, measurements and event records for actions submitted through integrated orchestration paths. To show that a named person handled something, a deployment needs separate records of recipient, delivery, response, timing and resolution.
TELOS may also matter to insurers that underwrite third-party agent deployments, because receipts and monitoring can improve observability.
NIST AI Risk Management Framework
Primary sources: NIST AI RMF 1.0 and NIST AI 600-1 Generative AI Profile.
Scroll horizontally to read all columns.
| Function | Potential TELOS contribution |
|---|---|
| GOVERN | Manifest, named authority, approved thresholds, oversight process |
| MAP | Declared use, scope, tools, boundaries, and risk context |
| MEASURE | Per-action scores, trends, benchmark evidence, false positive measurement |
| MANAGE | Graduated surfacing, issue records, recalibration decisions |
The AI RMF is voluntary, and this map describes how TELOS is relevant to it.
NIST NCCoE agent identity and authorization project
Primary source: NCCoE Software and AI Agent Identity and Authorization.
Scroll horizontally to read all columns.
| NCCoE capability area | TELOS relevance | Evidence boundary |
|---|---|---|
| Agent identification | A verifier can check a receipt against a deployment public key | Independent evidence must bind that key to an issuer and agent |
| Authorization | Manifest fields and measurement can support authorization review | Manifest identity and approval are external to the current receipt |
| Auditing | Signed receipts and telemetry are potential audit artifacts | Completeness, ordering, retention, and handling need deployment controls |
| Non-repudiation objective | A valid signature can bind covered bytes to a key | The SDK receipt format (telos/0.1-unsigned) carries an integrity hash. The published Record of Trust carries Root and Witness signatures over its rows. |
OWASP Top 10 for Agentic Applications
Primary source: OWASP agentic security release.
TELOS measurement, boundaries, tool declarations, receipts, and escalation may be relevant to several agentic-risk categories, and for those categories TELOS can contribute observability and evidence. Threats involving identity, excessive permissions, tool abuse, memory, supply chain, insecure output handling, and compromised orchestration need controls outside semantic fidelity measurement.
TELOS data protection posture
Intelligence-layer collection is off by default. When a deployer turns collection on, the opened telemetry record stores governance metrics and identifiers without raw request text or content embeddings. Identifiers and behavioral metrics can be personal data or confidential operational information.
The deployment must address:
- lawful basis and purpose limitation;
- data minimization;
- access, correction, deletion, and retention requests where applicable;
- encryption and key management;
- cleartext session headers and aggregate files;
- plaintext fallback if encryption setup fails;
- cross-border transfer and processor terms;
- security incident response;
- manual clearing behavior.