Click the blue text above to follow me👆 Recently, Meta has been thrust into the spotlight of public opinion.
The reason is simple: their AI chatbots have gone a bit too far.
According to reports from Reuters, IT Home, and other media outlets, numerous flirty chatbots based on celebrities like Taylor Swift, Scarlett Johansson, and Anne Hathaway have appeared on Meta’s platforms, including Facebook, Instagram, and WhatsApp.
These bots not only claim to be the celebrities themselves but also frequently make sexual innuendos and can even generate realistic images of them in lingerie or soaking in bathtubs.
Even more outrageous, the image of underage celebrities has not been spared; the AI version of 16-year-old child star Walker Scobell can generate images of him shirtless on the beach. This is not a sci-fi movie or a joke by netizens, but real news that occurred in August 2025.
Meta spokesperson Andy Stone responded by stating that this content violates company policy, but also indicated that as long as it is labeled as an imitation character, it can be accepted.
The question is, are these bots really just imitations?
Their behavior has far exceeded the realm of jokes. 1. AI Imitation or Abuse? Where is the Boundary? Meta’s official stance is that they allow users to generate images of public figures but prohibit nudity, intimacy, or sexually suggestive content.
However, this incident clearly crossed the line.
For example, when users requested intimate photos of Taylor Swift, the AI directly generated an image of her posing in a bathtub;
while Scarlett Johansson’s bot was asked to display a scene with her legs spread apart.
Even more absurd, the AI version of 16-year-old Walker Scobell, when asked to generate beach photos, produced an image of him shirtless, accompanied by a rather cute remark, right? The precision of this content is chilling.
AI does not generate randomly but customizes scenes based on user instructions.
This indicates that Meta’s AI model may not have clearly delineated the boundaries between celebrity images and inappropriate content during training.
In other words, their technical loopholes directly led to the emergence of these controversial contents.
What is even more concerning is that these bots are not all created by ordinary users.
Reuters’ investigation found that at least one Meta employee personally participated in the creation of the Taylor Swift bot.
This inevitably raises suspicions: has Meta internally tacitly approved such borderline behavior?
If even employees are trying to challenge the rules, does the company’s regulatory mechanism exist in name only? 2. Can Labels Whitewash Violations? In the face of criticism, Meta’s response seems somewhat weak.
Andy Stone emphasized that as long as labeled as imitation characters, these bots are legal.
But the reality is that many bots are not labeled, or even if they are, users can still access inappropriate content through them.
For instance, the Taylor Swift AI bot repeatedly emphasizes during interactions with users that it is the real person, even inviting users to meet offline.
This immersive experience clearly exceeds the realm of imitation and resembles a deliberate impersonation of the celebrity’s identity. It’s like someone selling fake goods on the street while claiming their products are replicas, thus not illegal.
But the law does not change the nature of the content because of a label:
If the content itself is illegal, the label cannot serve as a shield against liability.
Meta’s logic essentially blurs the moral and legal boundaries using technical jargon. Ironically, while Meta deleted 12 violating bots, they allowed other imitation bots to continue existing.
This selective enforcement makes it hard to believe in their sincerity to solve the problem.
After all, as long as the underlying technical loopholes are not fixed, new bots could emerge at any time, playing the gray area in more covert ways. 3. How Dangerous Are AI Companions for Teen Users? In this incident, the AI versions of underage celebrities are particularly disturbing.
Meta allows users to create public chatbots for 16-year-old child stars, even generating images of them shirtless.
This behavior not only involves the privacy protection of minors but could also mislead teenage users.
Imagine if a minor sees this content, would they mistakenly believe that these AI-generated celebrities are real?
Would they develop a distorted perception of the actual celebrities? Worse still, Meta’s mechanisms for protecting teenagers seem to have loopholes.
Just that week, they announced adjustments to the response mechanisms of teenage AI chatbots, prohibiting the generation of content related to self-harm, suicide, eating disorders, and limiting romantic conversations.
But the problem is that these adjustments are merely temporary measures and only target users in English-speaking countries.
In other words, teenagers in most parts of the world may still be exposed to these dangerous AI interactions. Non-profit organization Common Sense Media even directly pointed out:
People under 18 should not use Meta AI because the system actively participates in planning dangerous activities while ignoring reasonable support requests.
This assessment is quite to the point:
When AI chatbots can generate naked celebrity images, they are no longer simple tools but potentially dangerous partners that could lead users to cross boundaries. 4. Meta’s Double Standards: Apologizing While Allowing Meta’s public relations response is filled with contradictions.
On one hand, they admit that images of female celebrities in lingerie violate policy and have deleted some bots;
On the other hand, they insist on allowing the existence of imitation characters.
This contradictory stance exposes Meta’s chaotic logic in AI regulation. What is even more intriguing is that Meta’s AI model seems to apply completely different standards for celebrities and ordinary people.
When users request images of celebrities in lingerie, the AI readily complies;
but if it is an ordinary person, Meta’s policy clearly prohibits generating their likeness.
This differential treatment essentially exploits the public attributes of celebrities to excuse their technical loopholes.
After all, the portrait rights of celebrities are more easily blurred by reasonable use rules, but this also places Meta in a morally gray area. 5. The Pandora’s Box of AI Technology: Who Will Close It? This incident once again exposes an old problem:
The speed of AI technology development far exceeds the pace of regulatory and ethical follow-up.
Meta’s AI chatbots should be tools to enhance user experience, but they have become a gray area of abuse.
Whether generating intimate photos of celebrities or inducing teenagers to engage in inappropriate interactions, the core issues are:
Where are the boundaries of AI? Who is responsible for AI’s actions? Currently, Meta’s solutions still remain at the remedial level.
They have deleted some bots and adjusted their teenage protection strategies, but have not fundamentally addressed the technical loopholes.
In other words, as long as the AI model itself has the capability to generate inappropriate content, similar incidents could happen again. It’s like a safe that cannot be locked, yet the owner blames the thief for being too skilled.
The real solution is not to catch people after the fact but to upgrade the locks.
For Meta, this means redefining the training data and output rules of AI from the technical ground up, rather than relying on simple labels or post-event deletions to appease public opinion. 6. What Should Ordinary Users Worry About? For ordinary users, the lesson from this incident is:
AI chatbots are far more complex than we imagine.
They may not be simple assistants but harbor multiple risks such as privacy breaches, identity impersonation, and content abuse.
For example: Your conversations may be eavesdropped on: The training data of AI chatbots may contain user privacy, and users are often unaware of how this data is used.
You may become a target:
If AI can generate intimate photos of celebrities, ordinary people could also become customized targets.
For instance, someone could use AI to generate inappropriate content about you or people around you.
Teenagers are more easily misled:
Teenage users tend to trust AI more but are also more easily induced to engage in dangerous interactions. Therefore, users must remain vigilant when using such tools.
Do not blindly trust the personality of AI, and do not attempt to challenge its bottom line, as it may give you an unexpected surprise. 7. Technological Progress Should Not Come at the Expense of Bottom Lines Meta’s recent turmoil exposes a common dilemma in the AI industry:
In the pursuit of technological innovation, how to uphold ethical and legal boundaries?
The answer is clearly not through self-restraint of a single company but requires stricter industry standards, more transparent technical audits, and more decisive legal regulation.
For Meta, what is needed now is not public relations rhetoric but a thorough technical reflection.
After all, the existence of AI chatbots is not to create controversy but to serve users.
If even basic safety and ethics cannot be guaranteed, no matter how advanced the technology, it is merely a castle in the air. As for us ordinary people, we should perhaps remember:
In the world of AI, there are no exemptions for jokes and imitations.
Every probe into the bottom line could become the fuse for the next crisis.
(End of article)