Maki People
HireAIScore rates Maki People 67 out of 100 (grade D, Concerning) under rubric v1.0, last reviewed September 8, 2026. Maki People ranks 1 of 82 vendors rated in Assessment AI, against a category average of 46. That score is drawn from 31 evidence items across 7 rubric criteria for Maki People — 25 with a cited source, 6 recording that nothing was located.
Some company details provided by Maki People · updated July 31, 2026
Score breakdown
Maki People, criterion by criterion
Each rubric criterion is scored out of 100 on the cited evidence, then weighted into the total.
Weighted contribution shown to the right of each bar.
Article 11 Technical Documentation
Weight 20% · category avg 50
64
+12.8
Bias Audit Transparency
Weight 18% · category avg 41
80
+14.4
FRIA Support
Weight 15% · category avg 31
62
+9.3
Data Governance Disclosure
Weight 15% · category avg 53
64
+9.6
Human Oversight Design
Weight 12% · category avg 55
66
+7.9
Post-Market Monitoring
Weight 12% · category avg 38
55
+6.6
Customer Documentation
Weight 8% · category avg 58
74
+5.9
Total weighted score
66.5(rounded to 67 for display)
Category avg is the mean raw score on that criterion across the 82 Assessment AI vendors in scope of this rubric, this one included.
Strongest criterion
Bias Audit Transparency
80raw / 100
Contributes 14.4 to the total. 80 against a 41 category average.
Weakest criterion
Post-Market Monitoring
55raw / 100
Contributes 6.6 to the total. 55 against a 38 category average.
Cited evidence
The evidence behind the score
Every score is backed by at least one cited piece of evidence. Open a criterion to see what was found, when, and where.
Download diligence recordEvidence ledger
- Items
- 31
- Documentation
- 21
- Audit report
- 2
- Public statement
- 2
- Absence
- 6
- With a source URL
- 25 of 31
- Source hostnames
- 7
- Fewest items
- 4
- Article 11 Technical Documentation
These figures measure how thoroughly Maki People was reviewed, not how Maki People performed — an absence row, recording that nothing was located, is counted like any other item.
Article 11 Technical Documentation
4 items64
- DocumentationCaptured Sep 8, 2026
Science page documents IRT modeling, DIF/item-bias and 80%-rule Adverse Impact checks, a 12-step assessment build process, AI-vs-human-rater benchmarking (ICC/ANOVA) and transcription accuracy metrics (WER/CER/BLEU), and BARS rubric-based explainable scoring described as avoiding 'black-box reasoning.'
- DocumentationCaptured Sep 8, 2026
Trust Center lists EU AI Act 'Instructions for Use', an 'EU AI Act - Information for MakiPeople customers' deck, an 'AI Governance Policy', an 'AI Logging and Record-Keeping Policy', an ISO 27001 Statement of Applicability, and a Technical documentation section (infrastructure/network diagram), provided to customers on request (access-gated).
- DocumentationCaptured Sep 8, 2026
Public Help Center article sets out the scoring mechanics: BARS anchors per sub-dimension, how sub-dimension scores aggregate into a skill score, CEFR criteria for language assessments, and the exact deterministic scoring formulas for single/multiple choice items.
https://help.makipeople.com/en/articles/180920-how-are-my-tests-scored
- AbsenceCaptured Sep 8, 2026
No full public Annex IV-aligned technical pack or model/system card is published and ISO 42001 is not held; the EU AI Act Instructions for Use and the deeper technical documentation remain access-gated rather than openly downloadable.
Bias Audit Transparency
4 items80
- Audit reportCaptured Jul 17, 2026
Publicly viewable Holistic AI NYC LL 144 bias audit of 'Mochi' (voice screening): 3,920-interview factorial design, standalone and intersectional impact ratios by sex and 7 EEO-1 race/ethnicity groups (≈0.85–1.00), decision-support framing. Maki dates its latest Holistic AI audit to March 2026.
- Audit reportCaptured Jul 17, 2026
Publicly viewable Holistic AI NYC LL 144 bias audit of 'Shiro' (BARS-based assessment engine), a second independent audit report linked from the Trust Center's AI section.
- DocumentationCaptured Sep 8, 2026
Security page states the assessments are 'audited annually by Holistic AI under NYC Local Law 144, across gender and race/ethnicity at standalone and intersectional levels' - a public audit-cadence commitment - and links both signed reports.
- DocumentationCaptured Jul 17, 2026
Science page corroborates the audit posture: DIF and Adverse Impact (80% rule) checks at build-time and continuously, with independent validation.
FRIA Support
4 items62
- DocumentationCaptured Sep 8, 2026
Trust Center's AI section lists 'FRIA (EU AI Act) - Information for MakiPeople Customers' and 'EU AI Act - Information for MakiPeople customers', both provided to customers on request (access-gated) - purpose-built deployer inputs for an Article 27 FRIA.
- DocumentationCaptured Jul 17, 2026
'DPIA - Summary for MakiPeople customers' published on the Trust Center (access-gated); its description states it gives customers summary information about Maki's latest Data Protection Impact Assessment for the hiring platform - further deployer-facing input for an Article 27 FRIA.
- Public statementCaptured Jul 17, 2026
Public Maki webinar, 'Using AI in HR: what you need to know about the EU AI Act,' aimed at helping deployers understand their EU AI Act obligations.
- AbsenceCaptured Sep 8, 2026
The FRIA deck, the EU AI Act customer deck, the DPIA summary and the Instructions for Use are all access-gated, so a deployer working only from Maki's public materials could not complete a FRIA without first requesting access.
Data Governance Disclosure
4 items64
- DocumentationCaptured Sep 8, 2026
Trust Center's Certifications section lists a SOC 2 Certificate alongside the ISO 27001 Certificate, ISO 27001 Statement of Applicability and Cyber Essentials Certificate; the SOC 2 report is served via 'Request access'. Maki states the report is dated 16 June 2026 and attested by Prescient Security.
- DocumentationCaptured Jul 17, 2026
Trust Center hosts a publicly viewable subprocessor list ('MakiPeople - List of personal data subprocessors', dated 2026-03-30) plus an AI Governance Policy, Data Management Policy and Cryptography Policy (access-gated).
- DocumentationCaptured Jul 17, 2026
Privacy policy defines controller/processor roles, a 2-year post-login retention period, candidate GDPR rights, and a privacy@makipeople.com contact.
- AbsenceCaptured Sep 8, 2026
Production-model training-data provenance, any protected-attribute exclusion list, and whether candidate data is used to train models are still not disclosed; ISO 42001 is not held and the SOC 2 report itself is access-gated rather than publicly verifiable.
Human Oversight Design
7 items66
- DocumentationCaptured Sep 8, 2026
Recruiter manual for overriding an AI-generated score: an override reason is required, the original AI score stays visible next to the override, the interface displays who made the change, every override is retained in a consultable history, dimension and overall scores recalculate, and the feature is role-gated (Interviewers excluded).
https://help.makipeople.com/en/articles/704447-overriding-an-ai-generated-score
- DocumentationCaptured Sep 8, 2026
Recruiter manual for interrupting the system: archiving an assessment suspends all ongoing candidate sessions instantly and blocks new instances while preserving scores and data, is reversible, is restricted to Hiring Manager/Recruiter/Administrator roles, and is expressly framed as Maki's implementation of EU AI Act Article 14(4)(e) interruptibility.
https://help.makipeople.com/en/articles/654167-pausing-an-assessment-when-an-issue-is-detected
- DocumentationCaptured Sep 8, 2026
Recruiter manual for Article 86 explanation requests: what AI does and does not do in each assessment type, instruction never to reject on an automated score alone, response templates, a 'Copy link to assessment report' flow for sharing the candidate-facing scorecard (explicitly not the hiring-manager scorecard), and a recommendation to record such requests.
- DocumentationCaptured Sep 8, 2026
Candidate-facing AI-use disclosure article, automatically included in the scorecard sent to every candidate: names each stage where AI is used (Mochi conversational interview, transcription, scoring, proctoring, feedback generation, recruiter guidance), states proctoring signals do not auto-disqualify, and sets out Article 86 rights.
https://help.makipeople.com/en/articles/659390-how-is-ai-used-in-my-maki-assessment
- DocumentationCaptured Sep 8, 2026
Security page states there is no automatic-rejection feature, that a Maki score can never automatically trigger disqualification, that any adverse decision requires human review of the full candidate results, and that a rejection is in principle a GDPR Article 22 decision requiring meaningful human intervention by someone with authority to change it.
- Public statementCaptured Sep 8, 2026
On-the-record vendor responses: 'human review of every adverse decision is clearly mandated in our EU AI Act Instructions for Use - this is a requirement, not a recommendation,' alongside an acknowledgement that Maki holds no centralised data on how often that review occurs, since customers run it in their own ATS.
- AbsenceCaptured Sep 8, 2026
No per-jurisdiction oversight restriction or confidence-threshold configuration is documented, and no override-rate or human-review telemetry is available to Maki or its customers.
Post-Market Monitoring
4 items55
- DocumentationCaptured Sep 8, 2026
Public Atlassian status page for Candidates and Recruiters components with 30-day uptime, incident history and subscriptions; operating since July 2023, though no incident has ever been posted and it is not linked from the main site.
- DocumentationCaptured Jul 17, 2026
Trust Center lists an Incident Management Policy, GDPR Breach Notification Procedure, Data Breach Procedure, an AI Logging and Record-Keeping Policy, and a Business Continuity/DR plan with a documented 2024 recovery test (access-gated) - a defined incident channel and model-change record-keeping.
- DocumentationCaptured Jul 17, 2026
Science page describes continuous monitoring for model drift and bias plus ongoing psychometric analyses by an in-house science team.
- AbsenceCaptured Sep 8, 2026
No public model-update changelog and no AI-specific incident-reporting channel; the incident-management and record-keeping policies remain access-gated, and the public status page carries no incident history.
Customer Documentation
4 items74
- DocumentationCaptured Sep 8, 2026
Public Help Center with separate recruiter and candidate tracks: step-by-step product manuals covering AI-score override, assessment interruption, Article 86 explanation requests, results interpretation and ATS integration, plus candidate articles on AI use and scoring that are attached to every candidate scorecard.
https://help.makipeople.com/en/articles/704447-overriding-an-ai-generated-score
- DocumentationCaptured Sep 8, 2026
Actively maintained Trust Center (Vanta) with EU AI Act Instructions for Use, security/privacy/AI policies, a DPIA customer summary, a FRIA information deck, a publicly viewable subprocessor list, and certifications across seven sections.
- DocumentationCaptured Sep 8, 2026
Dedicated Security & Trust page consolidates certifications, fairness and accessibility commitments, an explicit 'How hiring decisions are made' section, EU AI Act and LL 144 claims, and links to the Trust Center.
- AbsenceCaptured Sep 8, 2026
The EU AI Act Instructions for Use are access-gated rather than published; there is no standalone downloadable DPA (per Maki, it is appended to every Master Service Agreement by default); the NYC LL 144 deployer/compliance deck is customer-only, not public.
Editorial notes
Company overview
Maki (Maki SAS) is a Paris, France-based AI hiring company founded in 2021 by Maxime Legardez, Paul-Louis Caylar, and Benjamin Chino. Its core product is a science-backed pre-hire assessment and screening platform spanning hundreds of skills, voice screening in 45+ languages, and behavioral/situational evaluations, now wrapped in conversational AI agents that also cover interview support and candidate self-scheduling. The company raised a ~€27.8M Series A (led by Blossom Capital) in early 2025, reports rapid enterprise growth, and counts Fortune 2000 clients such as H&M, BNP Paribas, Capgemini, PwC and Deloitte. Pricing is credit-based with custom enterprise quotes, placing it firmly in the enterprise tier.
Regulatory exposure
Maki has heavy exposure to all major AI-hiring regimes. As an assessment/screening tool that scores candidates, it is an EU AI Act Annex III high-risk system, and its Paris HQ plus large EU client base put it squarely under EU jurisdiction; it also markets to US employers, triggering NYC Local Law 144 AEDT bias-audit duties and emerging Illinois/Colorado obligations. Maki backs its 'EU AI Act alignment' and 'NYC Local Law 144 bias-audited assessments' claims with two publicly viewable, independent 2026 Holistic AI LL 144 audits - for its Shiro assessment engine and its Mochi voice-screening tool - each disclosing the AEDT description, methodology, and full impact-ratio tables, alongside ongoing DIF and Adverse Impact (80% rule) testing. Human oversight is its clearest strength: Maki publishes recruiter manuals documenting a score-override control that requires a written reason and retains a full, attributed override history, and an assessment-interruption control mapped explicitly to Article 14(4)(e), alongside a candidate-facing AI-use disclosure attached to every scorecard. Documentation depth still trails the strongest peers: it holds no ISO 42001, publishes no full Annex IV technical pack or model card, and much of its Trust Center (Instructions for Use, the DPIA summary, the FRIA deck, the SOC 2 report) is available to customers on request rather than openly downloadable.
Path to a higher score
With independent LL 144 audits, a SOC 2 report, a populated Vanta trust center and now a documented oversight workflow in place, the remaining gains are almost entirely about openness rather than existence. The single highest-leverage move is un-gating: the Instructions for Use, the FRIA and EU AI Act customer decks, the DPIA summary and the SOC 2 report are all served behind a request form, and this rubric grades whether a deployer could work from public materials alone. Beyond that: publish a full Annex IV-aligned technical pack or model card rather than gating it; pursue ISO 42001; disclose production-model training-data provenance, a protected-attribute exclusion list, and whether candidate data trains models, which is what separates Maki from the strongest data-governance scores; carry the stated annual Holistic AI cadence into a second published cycle and complement synthetic-data testing with live applicant-flow results; and put a model-update changelog and an AI-specific incident channel beside the existing status page.
Regulatory frame
What applies to assessment AI.
Assessment is the most-regulated surface in AI hiring. NYC LL 144, EEOC and OFCCP frameworks, Illinois HB 3773, and EU AI Act Annex III §4 all apply, with overlap on bias-audit obligations.
Compare
Maki People against its nearest-scoring peers.
The closest scores in Assessment AI, side by side.
- Side-by-sideMaki People vs HiPeople61D
- Side-by-sideMaki People vs Alva Labs61D
- Side-by-sideMaki People vs HireVue58F
- Side-by-sideMaki People vs Fuel5058F
Ranking
Others rated in Assessment AI
Ranked 1 of 82 by weighted total under rubric v1.0. The ordering is arithmetic on the rubric and carries no view on which tool suits a given hiring process.
- #1Maki PeopleThis profile67D
- #2HiPeople61D
- #3Alva Labs61D
Conflicts of interest
No vendor pays for placement, scoring, or removal. Casework - the consulting firm that operates this directory - provides paid services to some vendors. Any active or recent (within 24 months) commercial relationship is disclosed on the affected vendor profile and the review is reassigned to an independent reviewer. See the full policy on About.
Casework has no commercial relationship with this vendor.