Lessons from Meta's AI-Layoff Lawsuit: AI Performance Reviews and Labor Risk in the Philippines

Using the lawsuit that alleges Meta used AI to select who would be laid off, this guide explains, from a practical standpoint, the labor-law and data-protection points Japanese firms expanding into the Philippines need to watch when they introduce AI-driven performance reviews.

Part 1: Why This Matters

Step 1: The Philippine Business Context (3 min)

At Meta, the major U.S. tech company, 26 employees have sued the company, alleging that it may have used an AI system to select who would be laid off. The core issue is whether people who had taken "legally protected leave" — such as maternity or parental leave, or medical leave — were rated unfavorably by the AI and disproportionately chosen for layoff.

This topic hits close to home for Japanese companies expanding into the Philippines. The Philippines is widely used as a base where Japanese firms entrust call centers, shared accounting services, IT operations, and similar work. On these front lines, it is common to measure each employee's workload and attendance in fine detail through systems. Combine that with AI-driven performance reviews, and someone whose workload was zero because they were on leave can automatically receive a low score, creating a risk of unintended, unfair outcomes.

In the Philippines, workers' rights are strongly protected by law, and the rules on dismissal are in some ways stricter than in Japan. If you bring in AI-driven evaluation without knowing about the supervising authorities — DOLE (the Department of Labor and Employment, the agency responsible for Philippine labor administration) and the NPC (the National Privacy Commission, which oversees the handling of personal data) — you may run into serious trouble.

Scene: at the Manila office In your Manila office, you open a document from headquarters describing a new policy: "measure employee productivity with AI and use it to inform staffing decisions." You share it with the local manager at the next desk: "This Meta lawsuit could apply to us, too. If an employee on parental leave automatically gets a lower AI score, what happens under Philippine labor law? There are a few things I want to check before we roll this out." This material is here to help you start that conversation.

Step 2: Key Points of the Source Article (5 min)

For study purposes, here are the facts reported in the source article, organized into a table.

Item	Details
Number of plaintiffs	26 Meta employees (anonymous)
Claim	An AI system was used to select layoff targets, and those who had taken medical, childcare, or family-care leave were treated unfavorably
Where the suit was filed	U.S. federal court in Oakland, California
Scale of the layoffs	8,000 employees, about 10% of the total (scheduled to begin May 2026)
Meta's stated purpose for the layoffs	To improve the company's efficiency and fund other investments (internal document, April 2026)
Methods allegedly used	Internal AI systems, keystroke and activity monitoring data, dashboards showing AI usage, assistance with ranking evaluations
What the 26 plaintiffs share	All had taken protected leave and requested / received disability accommodations
Current employment status	All 26 are still employed; the layoffs are scheduled to begin July 22, 2026
Partial breakdown	8 women who took maternity / pregnancy-related leave, 4 men who took parental leave, 1 woman who took family-care and bereavement leave
Laws cited as grounds	Family and Medical Leave Act, Americans with Disabilities Act, Pregnancy Discrimination Act, Pregnant Workers Fairness Act, and others
Meta's rebuttal	The claims are baseless; staffing decisions were made by people, not AI

Source: CBS News — "26 Meta workers sue over alleged AI-aided layoffs targeting employees on medical or family leave" (July 15, 2026)

This table was created from publicly reported facts for study purposes. Please check the original article at the link above for details.

Step 3: Comprehension Check (5 min)

Q1. How many employees sued Meta, and what are they claiming?

Hint: The number is 26. Pay attention to the connection with leave.

Q2. Roughly what percentage of Meta's total workforce do these layoffs represent?

Hint: The figure of 8,000 and the overall proportion are stated in the text.

Q3. According to the lawsuit, what data and mechanisms were allegedly used to select layoff targets? Try to name three.

Hint: These include keystroke monitoring and dashboards showing AI usage.

Q4. The 26 plaintiffs share something in common. What is it?

Hint: The two keys are leave and disability accommodation.

Q5. How is Meta responding to the lawsuit?

Hint: The answer lies in whether the decisions were made by "AI" or by "people."

Part 2: Putting It Into Practice

Step 4: Steps for Rollout in the Philippines (10 min)

Here is a five-stage approach to avoid trouble when introducing AI-driven performance reviews or productivity measurement at a Philippine site.

Stage	What to do	Philippine-specific caution
1. Clarify the purpose	Define in writing what the AI will measure and what it will be used for	Tying it directly to dismissal decisions carries high labor-law risk, so limit the use
2. Design how leave is handled	Decide up front on rules that exclude maternity, parental, and medical leave periods from evaluation	Philippine maternity leave runs up to 105 days and is generous; design on the assumption that workload is zero during leave
3. Obtain consent for data	Before collecting monitoring data, explain it to employees and obtain consent	Follow the data protection law under the NPC (National Privacy Commission) and provide a point of contact for opting out
4. Final review by a person	Never use AI results as-is; build in a procedure where a person always reviews them	Because verbal agreements are valued in the local culture, keep records in writing
5. Check with local experts	Consult Philippine labor and legal experts before rollout	DOLE (Department of Labor and Employment) dismissal rules are strict, so checking in advance pays off

In Stage 1, make clear whether the AI evaluation will stay as "reference information" or be used as "grounds for dismissal decisions." The broader the use, the greater the legal danger. Start with a narrow scope.

In Stage 2, decide first how to handle leave periods. Workload naturally drops during leave. If you include that period in the evaluation, only those who took leave are disadvantaged. This is exactly the point at issue in the Meta lawsuit.

Data consent in Stage 3 is especially important in the Philippines. Monitoring data counts as personal data. Skipping the explanation to employees risks running afoul of NPC rules. As a budget guide, it is reassuring to allow a few tens of thousands of pesos for an initial legal consultation.

In Stage 4, a person always reviews the rankings and scores the AI produces. Check whether workload dropped due to leave or health, and correct the evaluation if needed.

In Stage 5, have a lawyer or labor-and-social-insurance expert familiar with Philippine labor matters review it once before rollout. Compared with redoing things later, consulting in advance is cheap.

Step 5: Common Mistakes and Fixes (5 min)

Mistake 1: "Using AI evaluations directly for dismissal decisions"

This is the mistake of using the productivity scores or rankings the AI produces directly for dismissal decisions, without human review. Someone whose workload dropped while on leave automatically gets a low score, producing an unfair result.

Bad example: An employee who ranked low in the AI productivity ranking was made a layoff candidate as-is.

Good example: The AI results were treated only as reference, and a person confirmed changes in workload caused by leave or health before the final decision was made.

Mistake 2: "Forgetting to exclude leave periods from the evaluation"

This is the mistake of running the system with maternity or medical leave periods still included in the evaluation calculation. Those who took leave are recorded with less workload and end up disadvantaged. Because Philippine leave programs are generous, the impact is even larger than in Japan.

Bad example: All employees were compared over the same period, and three months of maternity leave were included in the productivity calculation as-is.

Good example: Maternity and medical leave periods were excluded from the calculation in advance, so people could be compared fairly using only the periods they actually worked.

Mistake 3: "Not explaining the collection of monitoring data to employees"

This is the mistake of collecting monitoring data — keystroke logs, activity histories — without informing employees. In the Philippines this risks breaching the data protection law and also damages trust.

Bad example: Thinking it was to measure productivity, the company started collecting PC activity logs without telling employees.

Good example: The company explained in advance which data would be collected and why, obtained consent, and then began collecting. It also set up a point of contact for those who wanted to opt out.

Part 3: Going Deeper

Algorithmic selection — a computer automatically choosing people or things according to predefined computational rules — is a method where, following rules decided in advance, the computer picks out people or items on its own. In Philippine call centers, a system may automatically produce a performance ranking based on the number of calls answered and handling times; when using such rankings for evaluation, it is important to pair them with human review.

Keystroke and activity monitoring — recording key inputs and screen operations — is a mechanism that records how many times an employee pressed keys and how long they used software to gauge how much they worked. As remote work spreads at Manila outsourcing sites, explaining this to employees and heeding the rules of the NPC (National Privacy Commission) are essential before collecting such records.

Disparate impact — bringing a biased disadvantage to a specific group even when a rule looks neutral — refers to a state where a rule that appears the same for everyone ends up weighing heavily on people in a particular situation. In the Philippines too, because women tend to take more childcare and family-care leave, including leave-period workload in the evaluation tends to disadvantage women, which calls for caution.

Reasonable accommodation — adjustments that let people with disabilities work — means arranging the way work is done or the environment, within reason, according to a disability or health condition. At Philippine sites as well, employees granted health-conscious working arrangements should not be uniformly disadvantaged by AI evaluation alone; the evaluation mechanism itself needs to build in such accommodation.

Protected leave — leave safeguarded as a right by law — refers to leave such as maternity, parental, or medical leave, for which the law forbids treating someone unfavorably because they took it. In the Philippines, maternity leave is generously set at up to 105 days, and not reflecting the lower workload of this period directly in evaluations is key to preventing labor trouble.

Step 7: Consider How It Applies to Your Company (10 min)

Does your company's AI performance review handle "leave" correctly?

If your company evaluates employees using AI or automated aggregation, check how maternity and medical leave periods are being calculated.

Something to consider: Are leave periods excluded from the evaluation? Or are they reflected in the score as-is as "zero workload"? Checking with real data can be eye-opening.

Next action: Within one week, ask your HR staff how leave periods are handled in the evaluation mechanism you currently use.

Do you have employee consent for collecting monitoring data?

Review whether you have explained the data you collect to measure productivity to employees and obtained their consent.

Something to consider: Do employees know what data is collected, for what purpose, and since when? In the Philippines, an explanation in line with NPC rules is required.

Next action: Make a list of the data you collect and confirm with legal or local experts whether your explanation to employees is sufficient.

Is your setup one where "a person makes the final decision"?

Check whether you are using AI results as-is, or whether a human review procedure is built in.

Something to consider: Who reviews the rankings and scores the AI produces, at what stage, and how? Is that procedure recorded in writing?

Next action: Summarize the human-review procedure in a single document and share it across the relevant departments.

Part 4: FAQ

Q1. Is AI-driven performance review banned by law in the Philippines?

The Philippines does not yet have a law that uniformly bans the use of AI itself. However, the data you collect is subject to the data protection law, and dismissal decisions are subject to the strict rules of DOLE (Department of Labor and Employment). Don't relax just because it isn't banned; it's important to check from both the existing labor law and the data protection law.

Q2. Can we use the evaluation mechanism our Japanese headquarters decided on directly in the Philippines?

Bringing it in as-is is risky. Philippine leave programs are generous — maternity leave runs up to 105 days — and the rules on dismissal differ from Japan. Using Japanese standards unchanged risks disadvantaging those who take leave. Adjust the handling of leave periods to match local labor law before using it.

Q3. Are we free to collect any monitoring data that measures employee workload?

You are not free to collect it. Keystroke logs and activity histories count as personal data and must be handled in line with the rules of the NPC (National Privacy Commission). Explain to employees what you collect and why, and begin only after obtaining consent. It is reassuring to also set up a point of contact for those who wish to opt out.

Q4. Is it a problem to use AI-selected layoff candidates directly for dismissal decisions?

There is major risk. This is exactly the point being contested in the Meta lawsuit, because people whose workload dropped due to leave or health tend to be disadvantaged. Keep AI results as reference only, and always have a person review them before deciding. We recommend keeping that procedure on record.

Q5. If trouble does arise, where can we turn in the Philippines?

DOLE (Department of Labor and Employment) is the point of contact for labor matters, and the NPC (National Privacy Commission) for the handling of personal data. That said, responding after a problem has occurred is difficult. The best preparation is to consult experts well versed in Philippine labor and legal matters before rollout.

Tips for Getting the Most Out of This (3 Tips)

Make it your top priority to check the setting that excludes "leave periods" from the evaluation mechanism The heart of this lawsuit is that lower workload during leave was reflected in evaluations. First, confirm how maternity and medical leave periods are calculated in your own evaluation mechanism. Fixing just this point can reduce much of the risk.

Always insert one stage of "human review" into AI results Meta counters that "it was people, not AI, who decided." Put another way, human involvement is what's being questioned. Don't use the rankings the AI produces as-is; always insert a step where a person reviews them. Keeping that record will also help with later explanations.

Explain monitoring-data collection to employees before you collect it Explaining after you've collected the data is too late and damages trust. Convey what you collect and why in advance, and begin only after obtaining consent. In the Philippines, this order also matters for staying in line with NPC rules.

Bonus: How to Use PH AI Works

PH AI Works is a solutions company that supports the use of AI and technology in the Philippines. On this topic — "AI-driven performance reviews and productivity measurement" — we can help you introduce it safely, in line with Philippine labor law and data protection rules and grounded in local realities.

As a next step, you can consult us on things like the following:

Checking whether your current evaluation mechanism fits Philippine leave programs and data protection rules
How to explain monitoring-data collection to employees and build a procedure for obtaining consent
Designing a setup where people review AI results, and organizing how records are kept

Please feel free to get in touch. Consultations are free.

Lessons from Meta's AI-Layoff Lawsuit: AI Performance Reviews and Labor Risk in the Philippines