Peru Regulation of Law 31814 (DS 115-2025-PCM)

Binding law — in force

Explain high-risk AI decisions that affect human rights, with the key criteria and factors, in accessible language (Peru, Regulation of Law 31814 Art. 25.3)

Regulation of Law 31814, Art. 25.3 (explanation of results of decisions affecting human rights) · official text · In force: applies since 10 Sep 2026; a further phase applies from 10 Sep 2027, 10 Sep 2028, 10 Sep 2029 · PE

Peru's Regulation of Law 31814 Art. 25.3 requires that, where a high-risk AI system takes decisions affecting human rights, the explanation of its results be guaranteed to the affected users through mechanisms that let them understand the criteria and key factors used in the automated decision, in language accessible to the user. For private developers and implementers it applies in phases by sector from 2026-09-10. Detect a model-driven adverse decision with no stored criteria or key factors.

Who it applies to

• Duty falls on: developer, deployer
• Systems covered: high risk, automated decision
• Sectors: healthcare, education, employment, lending, public services, essential services
• Developers and implementers in Peru of high-risk AI systems (Art. 24.1) whose decisions affect human rights. Same private-sector phase-in as Art. 25.1 (2026-09-10 to 2029-09-10). Which decisions 'impact human rights' is a question for counsel (review flag).
• Not covered:
  • Use of AI systems for personal purposes (Art. 4(a))
  • Use of AI systems for defence and national security, provided the principles of Art. 7 are met (Art. 4(b))
• Whether it applies depends on facts outside the code; a person has to decide.

resp = client.chat.completions.create(model=MODEL, messages=msgs)
if 'deny' in resp.choices[0].message.content.lower():
    application.status = 'denied'
    send_email(applicant.email, 'Your application was declined.')

a = Assessment.model_validate_json(resp.choices[0].message.content)   # decision, reason_codes
if a.decision == 'deny':
    decisions.insert(app_id=application.id, status='denied', reason_codes=a.reason_codes,
                     model=resp.model, input_hash=hashlib.sha256(payload).hexdigest())
    send_email(applicant.email, render('adverse_action_notice.txt',
        reasons=a.reason_codes,
        role_of_ai='An AI model assessed your application; a reviewer can change the outcome.',
        correct_data_url='/profile/data', appeal_url=f'/appeals/new?decision={application.id}'))

Standards that recommend the same control

• Significant AI decisions should be documented and contestable with remedies (CoE Framework Convention) (CoE AI Framework Convention (CETS 225) · CoE Framework Convention on AI (CETS 225), Article 14(2))
• Financial institutions should disclose AI use and explain AI-driven decisions on request (MAS FEAT) (MAS FEAT Principles · MAS FEAT Principles — Transparency, Principles 12-14)
• Health AI/LMMs should be transparent and documented before deployment (WHO) (WHO AI-for-Health (LMM) Guidance · WHO guidance on large multi-modal models (2024), Executive summary (Figure 1), section 1.2 and Box 1: the six consensus principles of WHO's 2021 guidance)

Related incidents

• UnitedHealth nH Predict claim-denial litigation (2023-11; alleged (not proven)). A class action filed in November 2023 alleges that UnitedHealth's nH Predict model had a 90% error rate, measured by denials reversed on appeal, while only about 0.2% of members appealed. UnitedHealth disputes the allegations; the litigation is ongoing. Source: STAT News · evidence grade: primary · cited by Explain adverse AI-assisted decisions and offer a way to contest them — everywhere

Rule id pe-ai-regulation.high-risk-decision-explanation · review status: primary source derived

Binding law — in force

Tell users in advance what a high-risk AI system is for, what it does and what decisions it can take (Peru, Regulation of Law 31814 Art. 25.1-25.2)

Regulation of Law 31814, Art. 25.1-25.2 (algorithmic transparency: prior, clear information to users; visible labelling) · official text · In force: applies since 10 Sep 2026; a further phase applies from 10 Sep 2027, 10 Sep 2028, 10 Sep 2029 · PE

Peru's Regulation of Law 31814 Art. 25.1 requires the developer or implementer of a high-risk AI system to establish mechanisms for algorithmic transparency that inform the user in advance, clearly and simply, of the system's purpose or use, its main functions and the type of decisions it can take; Art. 25.2 adds that these may include visible labelling informing users beforehand that the product, service or content operates on AI, with its main capabilities and limitations, except for internal administrative support with no direct impact on decisions affecting rights or services. For private developers and implementers the duty applies in phases by sector from 2026-09-10. Detect a model used in hiring, credit, health or admission decisions with no prior AI notice in the flow.

Who it applies to

• Duty falls on: developer, deployer
• Systems covered: high risk, consequential decision
• Sectors: healthcare, education, employment, lending, public services, critical infrastructure, essential services
• Developers and implementers in Peru of AI systems in high-risk uses listed in Art. 24.1: management of critical national assets supporting essential services; evaluation of children and adolescents in education; selection, evaluation, hiring and dismissal of workers; access to social programmes; credit evaluation (fraud detection excepted); access to health services; screening, diagnosis or prognosis affecting health, emergency prioritisation and processing of sensitive data in electronic health records; emotion inference at work or in education (medical or safety reasons excepted); and other uses risking life, safety, rights, freedom or dignity with high risk of discrimination or high complexity for human supervision. Private developers and implementers apply Art. 25 in phases (Primera DCF): from 2026-09-10 for health, education, justice, security, economy and finance; 2027-09-10 for transport, commerce and labour and for small enterprises; 2028-09-10 for production, agriculture, energy and mining and for micro-enterprises; 2029-09-10 for all other uses. Which sector a use belongs to, and whether a use is high-risk, are questions for counsel (review flag; the SGTD answers consultations, Art. 24.2).
• Not covered:
  • Use of AI systems for personal purposes (Art. 4(a))
  • Use of AI systems for defence and national security, provided the principles of Art. 7 are met (Art. 4(b))
  • Visible labelling is not needed where AI supports internal administrative processes with no direct impact on decisions affecting rights or services (Art. 25.2)
• Whether it applies depends on facts outside the code; a person has to decide.

@app.post('/applications')
def apply(app_in: Application):
    resp = client.chat.completions.create(model=MODEL, messages=underwriting_prompt(app_in))
    return {'decision': underwrite(resp.choices[0].message.content)}

@app.post('/applications')
def apply(app_in: Application):
    notice = send_admt_notice(app_in.applicant_id, template='ai_decision_notice_v2')
    resp = client.chat.completions.create(model=MODEL, messages=underwriting_prompt(app_in))
    decision = underwrite(resp.choices[0].message.content)
    db.decisions.insert(app_in.id, decision, notice_id=notice.id)
    return {'decision': decision, 'ai_notice': notice.text}

Standards that recommend the same control

• Financial AI decisions should offer enquiry, appeal, and review channels to affected people (MAS FEAT) (MAS FEAT Principles · MAS FEAT Principles — Accountability (External), Principles 10-11)

Rule id pe-ai-regulation.high-risk-prior-information-and-label · review status: primary source derived

Binding law — in force

Keep a record of how a high-risk AI system works and people able to stop, correct or invalidate its decisions (Peru, Regulation of Law 31814 Art. 31.1, 31.4)

Regulation of Law 31814, Art. 31.4 (human oversight able to stop, correct or invalidate high-risk decisions) · official text · In force: applies since 10 Sep 2026; a further phase applies from 10 Sep 2027, 10 Sep 2028, 10 Sep 2029 · PE

Peru's Regulation of Law 31814 Art. 31 requires developers and implementers of high-risk AI systems to keep an up-to-date, accessible, preventive record of the system's operating principles, data sources, algorithmic logic and expected social and ethical impacts (31.1), and to implement human oversight in decisions with significant impact in health, education, justice, finance and access to basic programmes and services, with staff trained not to be biased by the system's results and able to stop, correct or invalidate its decisions (31.4). For private developers and implementers it applies in phases by sector from 2026-09-10. Detect a model whose output sets a decision status with no review, override or stop path.

Who it applies to

• Duty falls on: developer, deployer
• Systems covered: high risk, consequential decision
• Sectors: healthcare, education, employment, lending, public services, essential services
• Developers and implementers in Peru of high-risk AI systems (Art. 24.1); human oversight for decisions with significant impact in health, education, justice, finance and access to basic programmes and services. Same private-sector phase-in as Art. 25 (Chapter II of Title VI, 2026-09-10 to 2029-09-10). Whether the record of 31.1 must be published or only kept accessible, and what training 31.4 requires, are questions for counsel (review flag).
• Not covered:
  • Use of AI systems for personal purposes (Art. 4(a))
  • Use of AI systems for defence and national security, provided the principles of Art. 7 are met (Art. 4(b))
• Whether it applies depends on facts outside the code; a person has to decide.

score = model.predict_proba([features])[0][1]
status = 'approved' if score > 0.7 else 'denied'
applications.save(app_id, status=status)

if settings.HIGH_IMPACT_AI_PAUSED:            # kill switch
    raise ServiceUnavailable('automated scoring paused')
score = model.predict_proba([features])[0][1]
proposed = 'approved' if score > 0.7 else 'denied'
decisions.insert(app_id=app_id, proposed=proposed, score=score,
                 reason_codes=top_reason_codes(features, k=3), model_version=MODEL_VERSION,
                 status='pending_review')
# applications.save runs from the reviewer's approve/override handler

Rule id pe-ai-regulation.high-risk-record-and-human-oversight · review status: primary source derived

Binding law — in force

Do not use AI to infer race, political opinions, union membership, religion or sex life from biometric data (Peru, Regulation of Law 31814 Art. 23.1(d))

Regulation of Law 31814, Art. 23.1(d) (prohibited: inferring sensitive traits from biometric data; discriminatory classification) · official text · In force: applies since 22 Jan 2026 · PE

Peru's Regulation of Law 31814 (Supreme Decree 115-2025-PCM) Art. 23.1(d) prohibits, as improper use, AI systems intended to analyse, classify or infer people's sensitive data from their biometric data to deduce or infer racial or ethnic origin, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation, and to evaluate or classify people or groups where that causes or contributes to discriminatory or disproportionate results that violate fundamental rights. Detect a face or voice pipeline that computes a race, religion, sexual-orientation or political label.

Who it applies to

• Duty falls on: developer, deployer
• Systems covered: prohibited
• Developers and implementers (Art. 6(b), (d)) of AI systems in Peru, public or private, intended to infer sensitive traits from biometric data, or to evaluate or classify people with discriminatory or disproportionate results. Applies with the Regulation's entry into force, computed as 2026-01-22 (90 business days after publication on 2025-09-09). The business-day count and whether the private-sector phase-in of the Primera DCF (which names only Art. 25 and Chapter II of Title VI) also delays Art. 23 are questions for counsel (review flag).
• Not covered:
  • Use of AI systems for personal purposes (Art. 4(a))
  • Use of AI systems for defence and national security, provided the principles of Art. 7 are met (Art. 4(b))
• Whether it applies depends on facts outside the code; a person has to decide.

result = DeepFace.analyze(img_path=photo, actions=['age', 'gender', 'race'])
profile.update(age=result[0]['age'], ethnicity=result[0]['dominant_race'])

result = DeepFace.analyze(img_path=photo, actions=['age'])   # no 'race' action
profile.update(age=result[0]['age'])

Rule id pe-ai-regulation.no-sensitive-trait-inference-from-biometrics · review status: primary source derived