What Meta's 'Muse Image' Reversal Teaches About Generative AI, Likeness, and Personal-Data Risk

Using Meta's decision to pull an AI image-generation feature as a starting point, we walk through — from a hands-on, practical angle — how Japanese companies that run social media and use generative AI in the Philippines should manage likeness and personal data, including how to work with the NPC.

Part 1: Why This Matters

Step 1: The Philippine Business Context (3 min)

The Philippines is a country where social-media use is especially heavy, even for Southeast Asia. Instagram and Facebook are used daily as a first point of contact for corporate PR, recruiting, and sales. It is not unusual for employees to post company-event photos from their own accounts. In other words, images on social media constantly contain both "company assets" and "individuals' faces."

In Meta's case, the problem was that the content of posts on public accounts could be used as AI source material without the person's permission. For Japanese companies based in the Philippines, this is not someone else's problem. If your local staff's photos, store interiors, or product packaging appear on a public account, the very same issue lands on your own company.

On top of that, the Philippines has the NPC (National Privacy Commission), which oversees the handling of personal information. Because collecting and using personal data requires the person's consent, judging by the Japan headquarters' instincts alone when running social media or adopting AI tools can trip you up.

[At the Manila office] Monday morning, an office in BGC. A Japanese manager pastes this article into Slack and calls over the local PR lead. "Our official account is set to public, right? If a similar feature shows up again, our staff's photos could be used as source material too. Could you put together a list this week of the social-media accounts we use internally and the people who appear in them?" The local lead replies: "Understood. But staff personal accounts also have plenty of company-event photos on them. We should probably set rules for those too."

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

We have organized the facts of the source article point by point.

Item	Details
What happened	Meta pulled a new feature of its "Muse Image" AI image-generation tool shortly after releasing it
The feature in question	A feature that let Meta AI chatbot users tag a public Instagram account and generate or edit AI images based on that account's posts
Timing	Muse Image launched on Tuesday, and the feature at issue was withdrawn a few days later
The problem	Instagram users were included by default, so anyone with a public account could have had their likeness used without their knowledge
Meta's explanation	It admitted it had gotten the call wrong and said the feature is no longer available. It said its intent had been to offer a tool that supports creativity and to let people choose whether their public content could be used
Groups that criticized it	The US actors' union SAG-AFTRA hailed the withdrawal as a win it had secured, and had already been urging users to take steps to protect their likeness
Another criticism	Privacy International, a London-based human-rights group, told the BBC that this reflects how AI companies treat people's images and data as raw material
What's next	Muse Image was limited to Instagram, but the feature was slated to expand to WhatsApp, Facebook, and Messenger, and an AI video tool is reportedly in development
Other	Meta has declined to comment further

Source: BBC News — "Meta pulls new AI image feature after days of backlash" (July 2026)

This table was compiled from publicly available facts for educational purposes. Please check the original article at the link above for details.

Step 3: Comprehension Check (5 min)

Q1. What kind of feature did Meta pull?

Hint: Recall the relationship between public Instagram accounts and the Meta AI chatbot.

Q2. What was the biggest reason users pushed back so strongly?

Hint: The key is whether users chose to opt in themselves or were included from the start.

Q3. How does Meta itself explain the withdrawal?

Hint: There are two parts — wording that admits fault in its own decision, and an explanation of its original intent.

Q4. Name two groups that criticized the feature.

Hint: One is a US actors' union; the other is a London-based human-rights group.

Q5. Which services was Muse Image said to be expanding to, and what other tool is reportedly in development?

Hint: Recall the names of Meta's messaging services. There is also a mention of video.

Part 2: Putting It Into Practice

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

Here is a way to keep using generative AI at work while preventing incidents involving likeness and personal data.

Step	What to do	Philippines-specific notes
1. Audit	List the social-media accounts your company manages and the people who appear in them	Look beyond official accounts — find accounts that branches or departments created on their own
2. Re-collect consent	Prepare consent forms that define how employees' and customers' photos may be used	Follow the NPC's approach to personal consent, and provide the forms in both English and Tagalog
3. Review settings	Check and adjust each service's visibility and AI-use settings	Because new features arrive without notice, build a quarterly re-check into someone's job
4. Set internal rules	Put in writing which materials may and may not be fed to AI tools	The habit of settling things verbally runs deep, so always keep a signed written record
5. Train and inform	Hold briefings for local staff and use cases to build understanding	As a rough budget, figure from a few thousand pesos per session, including materials, venue, and light refreshments. Estimate the actual amount internally

Step 1: Audit social-media accounts and the people in them First, surface every account connected to your company. Beyond the official accounts, you will often find ones that branches or sales teams run on their own. Then build a table of which employees and customers appear in each post.

Step 2: Re-collect consent A photo of someone's face counts as personal data. In the Philippines, the NPC oversees the handling of personal data, and broadening how you use it without the person's consent will cause problems. For photos already posted, explain the purpose again and have people sign a consent form. Providing the document in Tagalog as well as English builds more buy-in on the ground.

Step 3: Review each service's settings As this episode shows, new AI features can include you by default. First, consider whether a public account really needs to stay public. Then, if there is an option to control AI use, set it to exclude your content from training and generation. Because settings screens change with every update, decide who checks them and how often.

Step 4: Put internal rules in writing Write down where the line is on what may be fed to AI tools. Be specific — for example, do not enter employees' photos, customers' contact details, or the contents of contracts. In the Philippines, verbal arrangements tend to stick, but they are not handed over when staff change. A signed written record protects you later.

Step 5: A briefing for local staff Handing out rules does not change behavior. Work with your Manila lead to build briefing materials that fit local workflows. In the meeting, present this Meta case and have people discuss how they would feel if their own photo were used without permission — that deepens understanding. Always leave time for questions at the end.

Step 5: Common Mistakes and How to Avoid Them (5 min)

Mistake 1: Assuming "if an account is public, anyone can use it freely"

The idea that information is fair game for anyone just because it is public is dangerous. In this case, the criticism was precisely that a person's likeness could be used without permission even from a public account.

Don't: Grab a group photo of local staff from the official account and feed it straight into an AI tool to create a promotional image.

Do: Decide in advance which photos are approved, and feed only those for which the people shown have given written consent into the AI tool.

Mistake 2: Leaving default settings untouched

New features can include you even if you do nothing. The biggest trap is the assumption that "it doesn't concern me because I don't use it."

Don't: Skim the announcement of a new AI feature and keep operating without ever opening the settings screen.

Do: When you spot a new-feature announcement, open the settings screen that same week and check. Keep a record of the check as well.

Mistake 3: Deciding at headquarters alone, so it never reaches the field

Setting rules at the Japan headquarters means nothing if they never reach the field in the Philippines. When the language barrier and the habit of settling things verbally combine, the rules become an empty formality.

Don't: Drop a Japanese-language rules document in a shared folder and call it communicated.

Do: Prepare an English version and convey it at a briefing. Confirm understanding with a signature, and clearly note the responsible person and their contact details.

Part 3: Going Deeper

Generative AI AI that creates new text, images, and the like in response to a person's instructions. In the Philippines, a growing number of companies use it to support back-office work — for example, drafting replies to customer inquiries that mix English and Tagalog.

AI Image Generation Technology in which a computer creates new drawing- or photo-like images from a text description or a photo you provide. Muse Image falls into this category; for a retail shop in Manila, one use might be quickly preparing product-showcase images.

Likeness The identifiable features of a person, such as their face or figure. This is what the US actors' union SAG-AFTRA urged people to protect. In the Philippines too, when you use an employee's photo in advertising, you need to obtain the person's consent in writing.

Default Opt-in An arrangement in which a user is treated as having "agreed" from the outset, even without taking any action. Because Instagram users being placed in this state drew criticism, local subsidiaries in the Philippines should also make it a habit to check default settings before using a new service.

Personal Data Information that identifies a specific person, such as a name or a photo of their face. Because the NPC oversees its handling in the Philippines, you need to confirm internally the scope of permission before feeding a customer list or employee photos into an AI tool.

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

Surface the likeness risks hidden in your own social-media operations

Prompt: How many employees and customers appear on your official account? How far does the consent you have from them extend? Also check for accounts that branches run on their own.

How will you notice a platform's changes?

Prompt: Who inside your company first learned of this new feature? If no one noticed, you are missing a mechanism for catching such information. Decide who checks what, and how often.

Can local staff speak up with confidence?

Prompt: Is there an atmosphere in which a staff member who senses "this use might be risky" can say so frankly to a Japanese manager? In Philippine workplaces, a concern to avoid embarrassing others is at play, and worries can be hard to surface. A channel where people can raise issues anonymously helps.

Next action This week, list the social-media accounts your company manages and appoint one person to check the consent status of the people shown in them. Once the list is ready, holding a 30-minute joint review between the Japan headquarters and the local team will help you catch gaps in understanding early.

Part 4: FAQ

Q1. Once a photo is posted on a public account, is it beyond protection? There are ways to protect it. As here, a feature itself can be pulled in response to criticism. Start by reviewing your visibility settings. For accounts that must stay public for business reasons, a realistic approach is to keep written consent from the people shown and to limit the number of photos you release externally.

Q2. In the Philippines, how far can you use employees' photos? The NPC oversees the handling of personal data, and the person's consent is the baseline. A single line in the onboarding paperwork may not be enough. Record the scope of consent separately by purpose — whether for advertising or for internal materials only.

Q3. Can you apply the Japan headquarters' rules to the Philippine entity as-is? As-is, they fall short. Japan's legal system and the Philippines' are different, and the language on the ground is mainly English and Tagalog. We recommend using the headquarters' rules as a foundation while building a Philippines-specific version together with your local lead.

Q4. Should you stop adopting AI tools? There is no need to stop. What became a problem was material being used without the person's knowledge. If you separate the materials that may and may not be entered and keep records, you can very much continue using them at work.

Q5. How much budget should you set aside? It varies widely with scale, so we cannot give a single figure. The reliable approach is to estimate in pesos item by item — preparing consent forms, translation, and holding internal briefings. Start by counting how many people are affected in your company.

Tips for Making This Work (3 Tips)

This week, list your company's social-media accounts A conversation about risk cannot move forward until the scope is clear. Simply building a table with three columns — account name, visibility, and the people shown — makes the discussion concrete.

Decide who checks new-feature announcements, and how often This feature included users even when they took no action. If you decide as a matter of routine who opens the settings screen and when, you can move quickly the next time something similar happens.

Build the rules together with local staff Handing out a Japanese-language document will not move the field. Prepare an English version and convey it at a briefing with concrete examples. Confirming understanding with a signature helps a habit take root of not leaving things at the verbal stage.

Bonus: How to Work With PH AI Works

PH AI Works is a firm that supports the adoption of AI and technology in the Philippines. We can advise you with both the business use of generative AI and the practicalities of personal-data protection in the Philippines in mind.

As a next step, you can consult us on matters such as:

How to run a risk audit of your social-media operations and AI-tool use
How to prepare consent forms and internal rules for local staff in English and Tagalog
Where to draw the line on materials, and how to keep records, when bringing generative AI into your operations

Consultations are free, so please feel free to get in touch.

What Meta's 'Muse Image' Reversal Teaches About Generative AI, Likeness, and Personal-Data Risk