AI Standards

EU Artificial Intelligence Act

The EU AI Act classifies AI systems by risk: prohibited practices, high-risk systems, limited-risk systems with transparency obligations, and minimal-risk systems. Calibrate sits in the limited-risk category and is designed to remain there.

We do not infer emotions, do not categorise individuals by sensitive attributes, do not perform social scoring, and do not generate decisions about employment, education access, or credentialing. Calibrate evaluates observable coaching behaviours against the ICF competency markers — it does not evaluate the coach as a person, nor the client.

We have designed the product to fall outside Annex III high-risk categories by purpose-locking its use. Our Terms of Service explicitly restrict Calibrate from being used as the sole or determining factor in credentialing decisions, hiring, firing, promotion, or admission to training programmes. Calibrate is a formative and developmental tool. It augments a human assessor; it does not replace one.

In line with the Article 50 transparency obligations — which apply from 2 August 2026, with a transition period until 2 December 2026 for AI-content marking under the Digital Omnibus (Regulation (EU) 2026/1744) — every Calibrate report opens with a clear statement that the content is AI-generated and is not a substitute for human assessment for credentialing purposes. Calibrate already meets this obligation today.

ICF AI Coaching Framework and Standards (V1.01, 2024)

In October 2024 the International Coaching Federation published Version 1.01 of the AI Coaching Framework and Standards — the first formal ICF standards document for AI applications in coaching. The framework classifies AI coaching applications into four types: Scheduling, Data Processing, Interactive, and Conversational.

Calibrate is a Data Processing application, as defined by the ICF framework. We process data provided by the user — coaching session recordings — and generate structured outputs that a coach, mentor coach, or supervisor then uses in their professional practice. We do not deliver coaching services. We do not interact with the coachee. We do not replace human judgement. An independent panel of five ICF MCC credentialed coaches is being established to review and recalibrate Calibrate's evaluation logic on a semi-annual basis.

The ICF standard operates on a self-certification model. Below is Calibrate's self-declared conformance against the elements applicable to Data Processing applications.

Foundation — AI Ethics

• A.1.3 Algorithm Transparency — We publish documentation describing the types of algorithms used (large language models from Anthropic and Google) and their basic functionality.
• A.1.4 Action Explainability — Every evaluation report includes a section explaining how marker scores are reached, with verbatim evidence and reasoning. The default-to-Partial logic is documented openly.
• A.1.5 Bias — We disclose known potential sources of bias — including LLM training data limitations, language coverage, and the cultural assumptions embedded in ICF marker structures — in our documentation.
• A.1.6 Data Transparency — We publish a clear statement identifying who within Calibrate has access to coach and client data, and where the data is stored.

Co-Creating the Relationship

• B.4.1 Information on AI Design and Intent — We disclose the creators of Calibrate (an ICF MCC assessor and trainer), the theoretical basis (the ICF Core Competencies and PCC/MCC markers, including the 2026 SolutionsAcademy BARS), and the ownership structure (Calibrate is operated by Leading srl STP, established in Italy).
• B.4.3 Domain Specificity — Calibrate is explicitly domain-specific. It evaluates coaching sessions against ICF competency markers only. It does not generalise to other domains.

Assurance and Testing

• E.10.2 Validation of System Capabilities — Calibrate's outputs are continuously validated against official ICF Master Evidence Sheets. Periodic review by the MCC assessor panel (in formation) will provide a second, independent validation layer on a semi-annual cycle.
• E.10.3 Accuracy and Bias Monitoring — We monitor accuracy at the marker level and publish known false-positive and false-negative rates as our calibration work progresses.
• E.10.5 Training Data Monitoring — Calibrate uses calibration anchors at ACC, PCC, and MCC levels drawn from real evaluated sessions assessed by credentialed ICF assessors. The MCC panel, once operational, will review and update these anchors on a semi-annual basis.
• E.10.6 System Design Review by Coaching Experts — Calibrate is designed by an ICF MCC assessor and trainer. An independent panel of five ICF MCC credentialed coaches is being established to review the system design on a semi-annual basis. Until the panel is fully constituted, Calibrate meets the Basic requirement (review by at least one coaching expert).

Technical Factors

• F.12.1 Data Encryption — All data is encrypted in transit and at rest using industry-standard protocols.
• F.12.2 Secure Sign-On — Access to the system requires authenticated sign-on.
• F.13.1 Data Management Disclosure — Before any session is uploaded, users are informed about how the data will be processed, stored, and retained.
• F.13.2 Consent for Data Processing — Explicit consent is required before processing any session content. A three-way consent template — coach, client, and sponsoring organisation where applicable — is provided.

This conformance statement is reviewed every six months and updated as the standard evolves.

ICF Coaching Platform Standards (April 2026)

In April 2026 the International Coaching Federation published the Coaching Platform Standards, a complementary framework built around four pillars: capability, safety, governance, and service quality. Calibrate is not a full coaching platform — we do not provide matching, scheduling, or session management — but we operate inside the same ecosystem, and we hold ourselves to the same four pillars.

On capability, Calibrate's evaluation logic is calibrated against real coaching sessions assessed by ICF MCC assessors. Our marker structures cover ACC, PCC, and MCC credential levels.

On safety, every report includes a dedicated "Limits of this evaluation" section that names what the system can and cannot reliably assess, and reminds the reader that credentialing decisions require human assessment.

On governance, Calibrate is operated by Leading srl STP, established in Italy, with a Data Processing Agreement available to all organisational users, GDPR-compliant data handling, and a clear separation between data controller (Leading srl STP) and data processors (Anthropic, Google). API providers do not use Calibrate data to train their models.

On service quality, our evaluation methodology is documented, our marker logic is auditable, and our reports are designed to be discussed — between coach and mentor coach, between assessor and candidate, between supervisor and supervisee — not consumed silently.

EMCC Digital Ethics Guidelines (2023)

The European Mentoring and Coaching Council published the Digital Ethics Guidelines in July 2023 — the first comprehensive framework for technology-enabled coaching practice. The Guidelines cover integrity, confidentiality, quality standards, and professional conduct. Calibrate is built to honour each of these.

Integrity means we do not claim more than the tool can do. Calibrate does not deliver coaching, does not replace a mentor coach, does not certify competency, and does not produce a verdict. It produces structured evidence and a developmental reflection.

Confidentiality means coaching session content is processed under strict data protection: audio is transcribed and discarded according to documented retention rules; transcripts and reports are accessible only to the authorised user; no content is used for model training; and a three-way consent flow — coach, client, and sponsoring organisation where applicable — is required before evaluation.

Quality standards means our work is checked against the published ICF marker structures, not against an opaque internal rubric. When discrepancies emerge between Calibrate's output and official ICF Master Evidence Sheets, we treat them as defects to diagnose and correct.

Professional conduct means we work within the boundaries of the coaching profession rather than around them. Calibrate is built by an ICF MCC assessor and trainer, in dialogue with the credentialing community, and is reviewed regularly by practitioners with no commercial interest in the product's growth.

What Calibrate is not

Calibrate is not a coach. It does not hold sessions, does not give advice, does not establish a coaching relationship, and does not interact with clients in real time.

Calibrate is not a credentialing body. It does not award or deny credentials, does not communicate with ICF or EMCC on a candidate's behalf, and its output has no formal weight in any official credentialing process.

Calibrate is not an HR tool. It is not designed for, and is contractually restricted from, use in hiring, firing, promotion, performance evaluation, or any decision affecting a person's employment or career progression.

Calibrate is a developmental mirror — designed to help coaches, mentor coaches, supervisors, and training providers see their work more clearly, with structured evidence rather than impression alone.

Frequently asked questions

Answers about how Calibrate works, data protection and GDPR, AI standards, professional ethics, and ICF's guidance on AI in mentor coaching are now consolidated in one place. See all FAQs →

Get in touch

For questions about Calibrate's compliance posture, data handling, or contractual restrictions on use, contact us at [email protected]. For our full Data Processing Agreement, Terms of Service, or Privacy Policy, see the links in the footer.

Last updated: May 17, 2026 · Next review: November 17, 2026