NPC Advisory 2024-04: AI & Data Privacy Philippines

Last materially reviewed: October 5, 2026

Direct Answer

NPC Advisory No. 2024-04, dated December 19, 2024, explains how the Philippine Data Privacy Act of 2012, its Implementing Rules and Regulations, and National Privacy Commission issuances apply when artificial-intelligence systems process personal data. It does not create a separate Philippine AI statute. Instead, it applies the existing privacy framework to AI development and deployment, including training and testing.

The practical message is simple: using AI does not suspend privacy law. Organizations still need a lawful basis, transparency, proportionality, appropriate security, accountability, and working mechanisms for data-subject rights.

Key Takeaways

The Data Privacy Act applies to AI. The NPC describes the DPA framework as principles-based and technology-neutral.
Training and testing count. Privacy obligations can arise before an AI system reaches production.
Publicly available personal data remains protected. Public availability does not automatically remove DPA obligations.
Lawful basis must be identified before processing. AI convenience is not itself a lawful basis.
Rights must remain usable. Organizations should design AI systems so data subjects can exercise applicable rights before, during and after development or deployment.
AI governance should be documented. Data sources, purposes, roles, security, retention, human oversight and risk assessments should be traceable.

Choose Your Next Step

Choose the situation closest to yours. These are practical starting points; they do not establish that a particular AI use is lawful.

What happened or what are you planning?	What can you do?	Where to go next
You want staff to put customer records into an AI assistant.	Identify the data, purpose and lawful basis; check retention, provider training and access before uploading. Use synthetic test data while those checks are incomplete.	Work through the AI vendor due-diligence checklist and the lawful-basis test below.
You are training or testing a model with personal data, including public profiles.	Map sources and data flows, assess necessity and risks, and document controls before use. Public availability does not remove privacy protection.	Complete a privacy impact assessment; check the Advisory, Sections 1–3.
AI screens applicants, blocks accounts or influences access to a service.	Check whether the system involves profiling or automated decision-making, its registration duties and the risk to the person. Design meaningful human review and a contest route where required.	Use the automated-decision safeguards below.
You think an organization used your personal data in AI unlawfully.	Save the notice, output, dates and correspondence; write to its data protection officer with a specific request. An AI error alone does not prove a DPA violation.	Follow the NPC complaint route and prior-notice requirements if the issue remains unresolved.

Primary Authority

The official source is the National Privacy Commission, NPC Advisory No. 2024-04, titled Guidelines on the Application of Republic Act No. 10173 or the Data Privacy Act of 2012, Its Implementing Rules and Regulations, and the Issuances of the Commission to Artificial Intelligence Systems Processing Personal Data.

The Advisory is part of the NPC’s interpretation and guidance framework for privacy compliance. It should be read together with the Data Privacy Act, its IRR, and other relevant NPC circulars and advisories.

Who Should Read NPC Advisory No. 2024-04?

businesses developing or deploying AI systems;
organizations using generative AI with employee or customer data;
software and SaaS providers offering AI functions;
personal information controllers using AI for profiling or decision-making;
personal information processors handling data for AI systems;
teams collecting or preparing datasets for model training or testing;
privacy officers, security teams, product managers and procurement teams evaluating AI vendors.

AI Development, Training and Testing

Privacy compliance does not begin only when the model is released to users. If personal data is collected, organized, labeled, cleaned, combined, embedded, used for fine-tuning, evaluated, tested or otherwise processed during the AI lifecycle, the DPA framework can apply at those stages.

Organizations should be able to identify:

what personal data enters the AI workflow;
where the data came from;
the purpose of processing;
the lawful basis relied on;
which entity is the personal information controller;
which vendors or processors are involved;
how long data and derived records are retained;
how rights requests can be handled;
what security and access controls protect the data.

Lawful Basis Comes Before AI Processing

Identify the basis before collecting, uploading, training or testing with personal data. Consent is one possible basis, not a universal requirement. Apply RA 10173, Sections 12 and 13, to the actual data and purpose:

Classify the data. Ordinary personal information is assessed under Section 12. Sensitive personal information and privileged information need an applicable Section 13 exception; an ordinary legitimate-interest claim is not enough.
State the specific purpose. Check whether the chosen statutory condition genuinely fits. Adding an AI clause to a contract does not by itself make every AI use necessary to perform it.
Document the analysis and limits. If relying on legitimate interests for ordinary personal information, record the purpose, necessity and balancing tests in NPC Circular 2023-07, Sections 3–8. If no basis fits, change the workflow before using the data.

Example: A support team wants AI summaries of customer emails. It should separate ordinary account details from health information or other sensitive content, remove unnecessary fields, check vendor access and training settings, and assess each purpose. A customer-service purpose does not automatically justify reusing the same records to train a general model.

What About Publicly Available Personal Data?

One of the most important practical points in the Advisory is that personal data does not automatically lose Data Privacy Act protection merely because it is publicly available. A public profile, website, directory, post or database can still contain personal data.

Before using public data in AI development, organizations should ask:

What is the exact purpose of collection?
Is the collection compatible with the context in which the data was made public?
What lawful basis applies?
Is the volume proportionate to the purpose?
Does the source contain sensitive personal information?
Can the data subject reasonably exercise applicable rights?
Does scraping or automated collection trigger additional NPC guidance?

For scraping-specific obligations, see Data Scraping in the Philippines: What the NPC’s 2026 Guidelines Require.

Transparency for AI Systems

Privacy notices and other disclosures should be meaningful enough for users to understand what personal data is used and why. Depending on the system, relevant information can include whether AI is involved, categories of personal data, data sources, purposes, recipients, retention, vendor roles, significant automated processing and ways to exercise rights.

Transparency should not be reduced to a vague statement that “AI may be used.” The disclosure should reflect the actual processing.

Data Accuracy, Fairness and Proportionality

AI can amplify bad data. Incorrect, outdated, incomplete or biased personal data can affect model outputs, profiling and decisions. Organizations should therefore build processes to evaluate data quality, relevance and proportionality rather than collecting every available field “just in case.”

This is particularly important when AI affects employment, credit, insurance, education, healthcare, fraud detection, identity decisions or access to services.

Data-Subject Rights and AI

The use of AI does not eliminate the rights recognized by the DPA. Depending on the processing, individuals may have rights to be informed, access data, object, rectify inaccuracies, seek qualifying erasure or blocking, obtain qualifying portable data, claim damages and file complaints.

Organizations should design systems and records so they can find the relevant data, understand its source, correct or restrict it when legally required, and explain the role of automated processing where applicable.

See Data Privacy Rights in the Philippines.

Automated Decision-Making and Profiling

Distinguish ordinary AI assistance from profiling and automated decisions. NPC Circular 2022-04, Section 5(A) requires registration in all instances of covered data-processing systems involving automated decision-making or profiling. A system does not escape that check merely because it is supplied by a vendor.

Where an automated decision poses significant risk to a person’s rights and freedoms, Advisory 2024-04, Section 2(B)(2)(a) calls for meaningful human intervention by people with the necessary competence and authority, and a mechanism to question and contest the decision.

Example: If a recruitment model rejects an applicant, a practical review route should let an authorized reviewer examine the relevant data, consider a correction and change the outcome when warranted. Merely asking an employee to approve the score without examining it does not provide meaningful review.

Before deployment, record the decision owner, review trigger, complaint channel and records needed to explain the outcome. Use the automated decision-making and profiling guide for the wider framework.

Privacy Impact Assessments

A PIA is not limited to AI projects labelled high-risk. NPC Circular 2023-06, Sections 4–7 says a PIA should be undertaken for every processing system of a controller or processor involving personal data, including off-the-shelf software. It must cover the data inventory, purposes and lawful bases, lifecycle, privacy principles and risks.

For an AI workflow, map prompts, uploaded files, training or retrieval data, outputs, logs, vendor access, retention and deletion. Address identified risks through controls, and reassess when the processing materially changes. The PIA is generally retained rather than routinely submitted to the NPC, but must be available when requested in investigations or compliance checks.

Use the Privacy Impact Assessment Philippines guide to prepare the assessment. Features without a lawful basis or incompatible with privacy principles should be switched off under Section 7 of the Circular.

AI Vendors and Processors

Outsourcing an AI system does not outsource accountability. A personal information controller should understand what the AI vendor receives, whether prompts or files are retained, whether data is used for provider training, where data is stored, which subprocessors are involved, and how deletion, security incidents and rights requests are handled.

Contracts should align with the DPA and its IRR rather than relying only on the vendor’s standard product terms. See the AI Vendor Due Diligence Checklist.

AI Privacy Compliance Checklist

Define the AI use case and business purpose.
Map all personal data used for development, testing and production.
Record data sources, including public or scraped sources.
Identify the PIC, PIP and vendor roles.
Document the lawful basis for each processing purpose.
Separate ordinary and sensitive personal information.
Conduct and update the PIA for the personal-data processing system, including off-the-shelf AI tools.
Apply data minimization and retention limits.
Implement access control, logging and security measures.
Document competent, authorized human review and a contest route for automated decisions posing significant risks.
Provide accurate privacy notices.
Create a process for rights requests.
Review vendor contracts and subprocessors.
Prepare incident and breach-response procedures.
Reassess the system when models, datasets, purposes or vendors materially change.

Common Mistakes

Assuming public data is automatically free for AI training.
Collecting entire datasets when only a small subset is necessary.
Using sensitive data without identifying a Section 13 condition.
Allowing employees to paste confidential or personal data into public AI tools without controls.
Treating a vendor’s privacy policy as a substitute for an outsourcing review.
Launching before documenting data sources and model risks.
Making significant decisions through opaque automation without checking the rules on automated processing.

Frequently Asked Questions

Is NPC Advisory No. 2024-04 a law?

It is an official NPC advisory applying the existing Data Privacy Act, its IRR and NPC issuances to AI systems processing personal data. It is not a separate comprehensive AI statute.

Does it apply to generative AI?

It can. If a generative-AI workflow processes personal data during development, training, testing, prompting, retrieval, output handling or deployment, the DPA framework may apply.

Can an employer use AI with employee data?

Potentially, but the employer must identify the lawful basis, provide appropriate transparency, secure the data, limit processing to legitimate purposes, manage vendors and respect applicable data-subject rights.

Does the Advisory ban AI?

No. It focuses on accountable and lawful processing of personal data when AI is used.

Related Cybercode Guides

Official Sources

Disclaimer

Important: This article provides general educational information about Philippine privacy and AI governance. It is not legal advice. AI processing should be assessed against the actual data, purpose, sector rules, contracts and current NPC issuances that apply to the organization.

NPC Advisory No. 2024-04: AI Systems and Data Privacy in the Philippines