Frontline Data in the Physical-AI Era: The Human Archive Case and What Japanese Firms in the Philippines Should Do

An Era When India's Gig Workers Train the World's Robots: Data Collection for Physical AI From the Human Archive Case, and Its Implications for the Philippines

From the case of Human Archive, a fast-growing Indian startup, we explain the latest trends in on-site data collection for physical AI and the practical points of data privacy law compliance that Japanese companies in the Philippines should keep in mind.

---

Part 1: Why This Matters

Step 1: The Philippine Business Context (3 min)

Frontline work — household services, restaurants, hotels — is a sector that moves on a large scale every day in the Philippines too. In Manila and Cebu, there is an abundance of workers handling housekeeping, delivery, cleaning, and hotel room cleaning, and data on how people move at these sites is becoming valuable material for the AI and robots of the future. The Indian startup Human Archive has built exactly such a mechanism to collect that data from frontline workers, and is trying to sell it to frontier AI (the group of companies advancing cutting-edge generative-AI research and development).

When Japanese companies expand into the Philippines and operate factories, BPOs (back-office centers run on an outsourcing basis), hotels, and restaurants, they will eventually face the question, "Who handles our own on-site data, and how?" The controversy Human Archive stirred up in India is a topic highly likely to arise in the Philippines in the near future too.

On a Monday morning at a Manila office, the operations manager of your local subsidiary calls out to you. "Last week, a frontline staffer came to me with a question: another company offered to have them wear a camera-equipped cap to collect data. How should we respond?" You open the article's key points on your smartphone, share them with your colleague, and begin to talk about how the same wave is coming to the Philippines too.

Step 2: Key Points From the Original Article (5 min)

Source: TechCrunch — "This startup is betting India's gig economy can train the world's robots" (May 26, 2026)

This table was created for learning purposes based on facts from public information. For details, please check the original article at the link above.

Step 3: Comprehension Check (5 min)

Q1. Besides footage, name two types of data Human Archive collects from frontline workers.

Hint: Check the "Additional equipment" row in the Step 2 table. It includes data related to touch and movement.

Q2. What is the baseline hourly wage Human Archive pays Indian workers, and what is the hourly range competitors pay?

Hint: Both dollars and rupees appear in the article. Focus on the converted dollar amount.

Q3. Name the two companies that declined to partner with Human Archive.

Hint: They are major Indian housekeeping services. One of their executives made a clearly negative remark on social media.

Q4. Which Indian law does Human Archive claim to comply with? And which government agency has begun an investigation?

Hint: A 2023 law on personal data protection, and the ministry in charge of electronics and information technology.

Q5. Name one of the lead investors (a representative investor) in the $8.2 million it raised.

Hint: A VC named Wing also appears in the interview.

---

Related: see How AI Helps Philippine Business Leaders Stay Competitive in 2026.

Part 2: Putting It Into Practice

Step 4: Implementation Steps in the Philippines (10 min)

If you are considering an effort to use on-site data for AI training in the Philippines, jumping straight to contracts or equipment orders will create major friction locally. We recommend proceeding in stages through the following five steps.

Step 5: Common Failures and Countermeasures (5 min)

Failure pattern 1: "Operating with nothing more than an English translation of Japan headquarters' consent form"

Bad example: You translate, as-is, the personal-information consent form used in Japan and have Philippine frontline staff sign it. The staff sign without understanding the content, and the NPC (National Privacy Commission) later points out that "the validity of the consent is doubtful."

Good example: In addition to the English version, prepare a Tagalog version of the consent form. Furthermore, always take time to explain verbally what is filmed, who it is given to, and that it can be withdrawn at any time.

Failure pattern 2: "Setting compensation directly from figures from India or other countries"

Bad example: Referring to the $1-per-hour rate in the original article, you conclude you can pay the same level in the Philippines. Dissatisfaction that "it's below minimum wage" spreads at the site and goes viral on social media.

Good example: Check the minimum wage for the NCR and each region in official DOLE (Department of Labor and Employment) information, and design appropriate compensation for the additional work. Set it at a peso-denominated level local staff can accept.

Failure pattern 3: "Leaving the storage location of data and the handling of overseas transfers vague"

Bad example: You send the collected footage to Japan headquarters' servers and operate without keeping any records at the local subsidiary. When the NPC asks "what is the basis and consent for the cross-border transfer," you are stuck for an answer.

Good example: When moving data overseas, in accordance with the cross-border transfer provisions of the Data Privacy Act, put in place the protection level of the destination and the contract terms. Keep records at the local subsidiary too, building a structure you can check at any time.

---

Related: see How AI Strategy Helps Philippine SMEs Avoid Costly Adoption Failures.

Part 3: Going Deeper

Step 6: Related Technical Terms (5 min)

Egocentric video data (first-person-view footage) is footage shot from a camera worn on the head or chest, close to one's own field of view. If a room-cleaning staffer at a Philippine hotel works while wearing a cap-mounted camera, "the hand movements of bed-making" and "the order of arranging amenities" are recorded as-is, and you can later show it to new hires as training material.

Physical AI (AI that moves in physical space) is AI that operates machines that pick things up or walk in the real world, not just inside a screen. If you picture a serving robot carrying dishes to a table at a Philippine restaurant, you can see it as exactly an example of physical AI entering the field.

An RGB-D camera (a camera that captures color and depth simultaneously) is a camera that, in addition to ordinary color footage, can also record "the distance to the subject" at the same time. If you film the work of sorting goods at a Manila warehouse with an RGB-D camera, you can have a robot learn even the sense of distance to the shelves, which helps when you later advance automation.

The DPDP Act (India's Digital Personal Data Protection Act) is a law enacted in India in 2023 that governs the handling of personal data. Because the Philippines also has a similar law (the Data Privacy Act, Republic Act No. 10173), problems that occur in India are instructive for the Philippines too, and serve as a basis for judgment when choosing a local partner company.

A motion-capture suit (a full-body suit that records movement) is a suit that finely records a person's joints and body movements with sensors. If you have a staff member in charge of a skilled assembly process in Philippine manufacturing wear the suit, you can turn their movements into data as-is and use it for training at overseas sites or for considering automation.

Step 7: Thinking About Applying It to Your Company (10 min)

Discuss the "ownership" of on-site data internally

Who owns the footage and work logs generated daily at Philippine sites? The company, the worker themselves, or the AI service provider?

Prompt: Consider whether, when a frontline staff member changes jobs, the data recording their movements should be carried over or deleted.

Next action: Hold an internal meeting of about an hour with legal and operations staff, and check how "ownership of data" is written in your current contracts.

How to ensure transparency in compensation and consent

When adding compensation for providing data, you need a mechanism that does not create the anxiety that "refusing might put me at a disadvantage."

Prompt: What, concretely, is an environment in which frontline staff can say "no" without feeling psychological pressure? An anonymous survey or a third-party consultation desk are possibilities.

Next action: Set up a consultation desk where frontline staff can voice opinions anonymously, and build a mechanism to report the monthly situation to senior management.

Whether your company should become a "seller of data" or a "user of data"

Whether you take the position of collecting and selling data like Human Archive, or the position of buying data to develop your own AI, greatly changes both the required investment and the legal responsibility.

Prompt: Starting a data-sales business far from your core business may affect trust in your core business. Calmly evaluate its fit with your core business.

Next action: At a management meeting, summarize on a single diagram and discuss where your company should be positioned three years from now — "data provider," "data user," "both," or "neither."

---

Part 4: FAQ

Q1. If I have frontline staff in the Philippines wear cameras, which law should I be most careful about first?

The Data Privacy Act (Republic Act No. 10173) takes top priority. Because footage includes personally identifiable information such as faces and name tags, design on the premise of treating it as sensitive personal information. Also check the rules of the NPC (National Privacy Commission), and proceed to prepare, as needed, to place a DPO (Data Protection Officer) within your company.

Q2. Is the $1-per-hour rate Human Archive offered in India a useful reference for the Philippines too?

It cannot be applied as-is. Because the NCR (Metro Manila) minimum daily wage hovers in the 600-peso range in the Philippines, design compensation for the additional work taking local levels into account. Using the latest DOLE (Department of Labor and Employment) guidelines as your standard, decide amounts together with local HR staff.

Q3. Can I send the collected data to Japan headquarters or to an AI research institution in a third country?

You can, but in accordance with the cross-border transfer provisions of the Data Privacy Act, you need to put in place the contract terms and the individual's consent. Put yourself in a state where you can explain to the NPC whether the destination country's protection level is adequate and whether the contract commits the parties to equivalent protection.

Q4. What do I do if a frontline staff member says "I don't want to be recorded"?

Respect that wish and have them continue their work as usual without filming. It is important not to create an atmosphere where "refusing affects your evaluation." Because deference to the employer easily comes out strongly in the Philippines, prepare an anonymous mechanism that does not inform supervisors of the names of those who declined.

Q5. Japan headquarters says, "Isn't the controversy happening in India irrelevant to the Philippines?" How should I explain it?

The Philippines also has personal-information protection laws and is a society with very strong viral reach on social media. A clash between a major company and a startup, like the one that happened in India, can fully happen in the Philippines too. Explaining it to headquarters as "considering a Philippine version now, of the debate that is ahead in India, leads to maintaining trust later" tends to get through.

---

Tips for Making It Work (3 Tips)

Design the on-site consent process with both multiple languages and verbal explanation

With a paper consent form alone, frontline staff often cannot fully understand the content. Prepare a Tagalog version and visual explanatory materials, and always take time for verbal explanation. You can then keep evidence with which to explain, if the NPC (National Privacy Commission) inquires, that "consent was substantively established."

Write the purpose of data use and the retention period concretely into the contract

Writing only "for AI training" is too broad in scope and causes disputes over interpretation later. Narrow the purpose, as in "for creating our own training materials and improving operations," and write even the retention period and deletion procedure into the contract. The less ambiguity, the more trust accumulates between the frontline and headquarters.

Build a mechanism to report the reality of data collection to senior management monthly

Once you start collecting data, problems inevitably arise at the site, such as "footage is being captured even in unanticipated situations" or "equipment is being used while broken." Build a mechanism to report, once a month, the number of collections, whether there are complaints, and the state of the equipment to senior management, putting yourself in a state where you can correct course early.

---

Bonus: How to Use PH AI Works

PH AI Works supports the adoption of AI and technology, tailored to local context, for Japanese companies that have expanded into the Philippines and for Japanese companies considering expansion. Regarding this theme — data collection for physical AI and the handling of on-site data — we can also advise based on regulations and culture unique to the Philippines.

As a next step, you can consult us in a free consultation on matters such as the following.

• Checking the alignment between the Philippines' data privacy law and your on-site operations
• Support designing the consent process for frontline staff and multilingual explanatory materials
• Planning how to advance the adoption of AI and robots, and prioritizing operational efficiency

Please feel free to contact us.

---

References and Sources

• TechCrunch — "This startup is betting India's gig economy can train the world's robots" (May 26, 2026)
• National Privacy Commission (NPC) — official website of the Philippine National Privacy Commission
• Department of Labor and Employment (DOLE) — official website of the Philippine Department of Labor and Employment