When Careful Work Becomes Too Easy to Steal
James Cowman
Facebook may survive as a platform while losing much of the careful work that once gave it value. AI misuse, image theft, weak context, careless reposting, and exhausting moderation may not empty the platform. They may do something quieter. They may persuade serious contributors to withdraw their best images, research, and archival work from public view.
Predicting the disappearance of Facebook is probably a mistake. Platforms that large rarely collapse all at once. They survive through habit, scale, advertising, nostalgia, and the simple fact that millions of people still use them casually. Survival, however, is not the same as quality. A platform can remain active while becoming less valuable. It can continue generating engagement while driving away many of the people who made it worth visiting in the first place. The danger is not that Facebook will become empty. It is that the platform will remain full while more of its content matters less.
Hollowing begins when serious contributors stop sharing their best work. A photographer no longer posts original images. A researcher keeps carefully sourced material offline. A curator stops uploading rare finds because the risk of losing control has become too great. A writer decides that a thoughtful essay does not belong in a space where it can be flattened into drive-by comments, misread, stolen, or fed into someone else’s content machine. The work itself does not necessarily end. It moves elsewhere, becomes private, or disappears from public view altogether. Someone concerned with preservation stops treating the public feed as an archive and begins seeing it as a place where archival integrity can be destroyed.
Social media platforms do not seem to understand how much careful work depends on trust. A contributor must believe that authorship will be respected, context will matter, and theft will not be rewarded by the same system that asks creators to keep posting. When that trust collapses, people do not always make a dramatic exit. Often, they stop giving the platform anything meaningful.
I reached that point after an image I had purchased, prepared, and shared was taken from my Facebook page, stripped of its watermark, colorized or digitally manipulated, and then reposted to my own page. I already knew people were taking images. I understood that anything posted publicly could be saved, cropped, altered, and recirculated. I am also familiar with what AI and other digital tools can do. None of this was theoretical to me. Knowing that something is possible is different from having it thrown back in your face in your own space. The incident made the problem immediate. The risk was no longer distant or abstract. It was attached to my own work and returned to the very community where I had shared it.
Once that becomes clear, withdrawal is a rational response. It does not necessarily mean abandoning writing, research, photography, or preservation. It means leaving the platform that made the work vulnerable.
Online communities are not built equally by everyone. Research by Ricardo Baeza-Yates and Diego Saez-Trumper on user-generated content found that a small percentage of active users produced half of the content on Facebook and Twitter. When one of those serious contributors pulls back, the loss is greater than a single person leaving. The community loses a disproportionate amount of its depth.
This is why “just ignore it” is such poor advice. That response treats the problem as though it were merely an unpleasant feeling that a contributor should learn to tolerate. The actual loss is structural. When people doing careful work decide that a platform is no longer worth the risk, the platform loses more than their posts. It loses their expertise, judgment, standards, archives, and willingness to continue building something that others consume.
Research on online harassment points in the same direction. Data & Society found that many American internet users censor what they post because they fear abuse or harassment. The Pew Research Center has also documented how many Americans have experienced online harassment. Bad behavior online does more than create isolated unpleasant moments. It changes what people are willing to say, show, publish, and preserve in public. A contributor does not have to experience harassment in its most extreme form before changing behavior. Constant theft, misrepresentation, false accusations, bad-faith comments, and the knowledge that anything posted may be treated as free material can be enough. Eventually, the cost of sharing exceeds the benefit.
Much of that cost remains invisible to the audience. People see what has been posted, but they do not see what has been withheld. They never see the images that were not uploaded, the research that stayed private, or the archival material that was deliberately kept offline. They do not know which thoughtful contributor decided that public exposure was no longer worth the consequences. A platform can therefore appear active while its unseen losses continue to accumulate. That is the quiet form of collapse: not a shutdown, a headline, or a mass announcement, but the gradual disappearance of the work that required the most care.
Generative AI intensifies the problem because it changes the nature of copying. In the past, someone could steal or repost an image. That was bad enough. Now the image can be colorized, smoothed, enlarged, reconstructed, stripped of context, or altered in ways that are not immediately obvious. Facial details may change. Textures can be invented. Expressions can be subtly rewritten. The altered version may then circulate as though it has the same relationship to reality as the original. The image has not simply been copied. It has been rewritten.
That creates a serious problem for anyone concerned with photographic integrity. A photograph is not merely a visual object. It is evidence connected to a source, a moment, a process, and a chain of custody. When photographs are treated as raw material for endless alteration, the distinction between documentation and invention begins to disappear. Careful viewers may no longer know exactly what they are seeing.
That difference matters even more in fandom and cultural memory. Public memory is often assembled from photographs, captions, repeated stories, and fragments of information. When the photographs are altered, and their context is removed, the memory built from them becomes less reliable. The public does not merely receive more versions of the past. It receives a past that is increasingly difficult to verify. AI misuse is therefore different from ordinary reposting. Reposting can remove credit. AI alteration can remove reality. Over time, the altered version may become the one people remember.
Artists have already recognized this danger. A 2024 survey of 459 artists by Juniper Lovato and colleagues found serious concern about generative AI, including the scraping of digital work without consent, uncertainty over ownership, and the possibility that AI-generated material could compete with human artists in the marketplace. These concerns are not limited to commercial illustrators or fine artists. They apply to anyone whose work can be absorbed, altered, imitated, or repurposed after being placed online.
The U.S. Copyright Office has also recognized the seriousness of the conflict. Its report on generative AI training examines the use of large amounts of data, including copyrighted material, and the unresolved legal and economic questions surrounding AI systems. The law may still be trying to catch up, but creators are already dealing with the practical consequences. They must decide whether public sharing still makes sense when control, attribution, and context can disappear almost instantly.
A platform does not need to ban serious work to drive it away. It only has to make that work feel unsafe, unrewarded, and unprotected. Theft must remain easy, context weak, moderation exhausting, and digital manipulation nearly frictionless. After enough experience with those conditions, people who care about standards make the obvious decision. They stop feeding the machine.
The result is not silence. It is noise. There will be more material, not less. Altered photographs, recycled posts, quick reactions, fake images, shallow arguments, and low-effort engagement can fill a feed indefinitely. Generative AI makes this oversupply easier by reducing the cost of producing endless visual and written material. A 2024 paper by Yukun Zhang on generative AI and content platforms argues that inexpensive AI production can increase the amount of available content, contribute to information overload, and reduce user satisfaction. The danger is not disappearance but dilution.
The feed remains busy while its center of value shifts toward material that is cheaper, thinner, and less trustworthy.
Individual bad actors are part of the problem, but the larger issue involves incentives. Platforms reward activity more consistently than they reward care. Posting, reacting, sharing, arguing, and returning all produce measurable engagement. A stolen image can create engagement. A fake image can create engagement. Even an argument over whether an image is real can create engagement. The platform may prohibit certain behavior in its written rules while its mechanical and emotional incentives continue favoring motion over integrity.
Serious work operates by another standard. It requires source checking, restraint, and judgment. Sometimes it requires the decision not to publish something. A healthy archive is not created by uploading everything available. It depends on knowing what an item is, where it came from, how it should be presented, and when it should be protected. That kind of work cannot thrive in a system that treats every item as interchangeable content. A photograph, rumor, meme, AI reconstruction, restored scan, cropped repost, and researched essay all enter the same feed. The platform flattens the distinctions between them and then expects the audience to determine authenticity in the comments. That is not preservation. It is a system designed for confusion.
A Pew Research Center expert report on the future of online discourse warned that uncivil and manipulative behavior could cause social media to divide into more regulated safe spaces and less controlled free-for-all zones. That prediction feels increasingly accurate. People concerned with standards will not necessarily stop working. They will search for smaller, safer, and more controlled places to put the work. For some, that may mean a personal website, book, private archive, newsletter, moderated community, or curated collection. It may mean publishing fewer images and writing more essays. It may mean using illustrations instead of real photographs. Public platforms may still serve a purpose, but increasingly they will function as signposts rather than homes.
For this reason, I no longer believe Facebook is a safe place for serious photographic sharing. It may remain useful for announcements, links, casual conversation, and material that carries little risk. For serious archival and source-sensitive photographic work, I no longer consider that risk acceptable.
The future may not involve a mass exodus. It may be quieter and more damaging. Valuable images will remain offline, essays will move to personal websites, and larger archival projects will be reserved for more controlled forms of publication. Facebook will no longer be treated as an archive. It will be treated as a place where archives can be stripped for parts. That is what the hollowing out of serious work looks like. It is not the end of Facebook, creativity, research, or preservation. It is the end of trusting Facebook with anything that matters.
Sources Consulted
James Cowman
Facebook may survive as a platform while losing much of the careful work that once gave it value. AI misuse, image theft, weak context, careless reposting, and exhausting moderation may not empty the platform. They may do something quieter. They may persuade serious contributors to withdraw their best images, research, and archival work from public view.
Predicting the disappearance of Facebook is probably a mistake. Platforms that large rarely collapse all at once. They survive through habit, scale, advertising, nostalgia, and the simple fact that millions of people still use them casually. Survival, however, is not the same as quality. A platform can remain active while becoming less valuable. It can continue generating engagement while driving away many of the people who made it worth visiting in the first place. The danger is not that Facebook will become empty. It is that the platform will remain full while more of its content matters less.
Hollowing begins when serious contributors stop sharing their best work. A photographer no longer posts original images. A researcher keeps carefully sourced material offline. A curator stops uploading rare finds because the risk of losing control has become too great. A writer decides that a thoughtful essay does not belong in a space where it can be flattened into drive-by comments, misread, stolen, or fed into someone else’s content machine. The work itself does not necessarily end. It moves elsewhere, becomes private, or disappears from public view altogether. Someone concerned with preservation stops treating the public feed as an archive and begins seeing it as a place where archival integrity can be destroyed.
Social media platforms do not seem to understand how much careful work depends on trust. A contributor must believe that authorship will be respected, context will matter, and theft will not be rewarded by the same system that asks creators to keep posting. When that trust collapses, people do not always make a dramatic exit. Often, they stop giving the platform anything meaningful.
I reached that point after an image I had purchased, prepared, and shared was taken from my Facebook page, stripped of its watermark, colorized or digitally manipulated, and then reposted to my own page. I already knew people were taking images. I understood that anything posted publicly could be saved, cropped, altered, and recirculated. I am also familiar with what AI and other digital tools can do. None of this was theoretical to me. Knowing that something is possible is different from having it thrown back in your face in your own space. The incident made the problem immediate. The risk was no longer distant or abstract. It was attached to my own work and returned to the very community where I had shared it.
Once that becomes clear, withdrawal is a rational response. It does not necessarily mean abandoning writing, research, photography, or preservation. It means leaving the platform that made the work vulnerable.
Online communities are not built equally by everyone. Research by Ricardo Baeza-Yates and Diego Saez-Trumper on user-generated content found that a small percentage of active users produced half of the content on Facebook and Twitter. When one of those serious contributors pulls back, the loss is greater than a single person leaving. The community loses a disproportionate amount of its depth.
This is why “just ignore it” is such poor advice. That response treats the problem as though it were merely an unpleasant feeling that a contributor should learn to tolerate. The actual loss is structural. When people doing careful work decide that a platform is no longer worth the risk, the platform loses more than their posts. It loses their expertise, judgment, standards, archives, and willingness to continue building something that others consume.
Research on online harassment points in the same direction. Data & Society found that many American internet users censor what they post because they fear abuse or harassment. The Pew Research Center has also documented how many Americans have experienced online harassment. Bad behavior online does more than create isolated unpleasant moments. It changes what people are willing to say, show, publish, and preserve in public. A contributor does not have to experience harassment in its most extreme form before changing behavior. Constant theft, misrepresentation, false accusations, bad-faith comments, and the knowledge that anything posted may be treated as free material can be enough. Eventually, the cost of sharing exceeds the benefit.
Much of that cost remains invisible to the audience. People see what has been posted, but they do not see what has been withheld. They never see the images that were not uploaded, the research that stayed private, or the archival material that was deliberately kept offline. They do not know which thoughtful contributor decided that public exposure was no longer worth the consequences. A platform can therefore appear active while its unseen losses continue to accumulate. That is the quiet form of collapse: not a shutdown, a headline, or a mass announcement, but the gradual disappearance of the work that required the most care.
Generative AI intensifies the problem because it changes the nature of copying. In the past, someone could steal or repost an image. That was bad enough. Now the image can be colorized, smoothed, enlarged, reconstructed, stripped of context, or altered in ways that are not immediately obvious. Facial details may change. Textures can be invented. Expressions can be subtly rewritten. The altered version may then circulate as though it has the same relationship to reality as the original. The image has not simply been copied. It has been rewritten.
That creates a serious problem for anyone concerned with photographic integrity. A photograph is not merely a visual object. It is evidence connected to a source, a moment, a process, and a chain of custody. When photographs are treated as raw material for endless alteration, the distinction between documentation and invention begins to disappear. Careful viewers may no longer know exactly what they are seeing.
That difference matters even more in fandom and cultural memory. Public memory is often assembled from photographs, captions, repeated stories, and fragments of information. When the photographs are altered, and their context is removed, the memory built from them becomes less reliable. The public does not merely receive more versions of the past. It receives a past that is increasingly difficult to verify. AI misuse is therefore different from ordinary reposting. Reposting can remove credit. AI alteration can remove reality. Over time, the altered version may become the one people remember.
Artists have already recognized this danger. A 2024 survey of 459 artists by Juniper Lovato and colleagues found serious concern about generative AI, including the scraping of digital work without consent, uncertainty over ownership, and the possibility that AI-generated material could compete with human artists in the marketplace. These concerns are not limited to commercial illustrators or fine artists. They apply to anyone whose work can be absorbed, altered, imitated, or repurposed after being placed online.
The U.S. Copyright Office has also recognized the seriousness of the conflict. Its report on generative AI training examines the use of large amounts of data, including copyrighted material, and the unresolved legal and economic questions surrounding AI systems. The law may still be trying to catch up, but creators are already dealing with the practical consequences. They must decide whether public sharing still makes sense when control, attribution, and context can disappear almost instantly.
A platform does not need to ban serious work to drive it away. It only has to make that work feel unsafe, unrewarded, and unprotected. Theft must remain easy, context weak, moderation exhausting, and digital manipulation nearly frictionless. After enough experience with those conditions, people who care about standards make the obvious decision. They stop feeding the machine.
The result is not silence. It is noise. There will be more material, not less. Altered photographs, recycled posts, quick reactions, fake images, shallow arguments, and low-effort engagement can fill a feed indefinitely. Generative AI makes this oversupply easier by reducing the cost of producing endless visual and written material. A 2024 paper by Yukun Zhang on generative AI and content platforms argues that inexpensive AI production can increase the amount of available content, contribute to information overload, and reduce user satisfaction. The danger is not disappearance but dilution.
The feed remains busy while its center of value shifts toward material that is cheaper, thinner, and less trustworthy.
Individual bad actors are part of the problem, but the larger issue involves incentives. Platforms reward activity more consistently than they reward care. Posting, reacting, sharing, arguing, and returning all produce measurable engagement. A stolen image can create engagement. A fake image can create engagement. Even an argument over whether an image is real can create engagement. The platform may prohibit certain behavior in its written rules while its mechanical and emotional incentives continue favoring motion over integrity.
Serious work operates by another standard. It requires source checking, restraint, and judgment. Sometimes it requires the decision not to publish something. A healthy archive is not created by uploading everything available. It depends on knowing what an item is, where it came from, how it should be presented, and when it should be protected. That kind of work cannot thrive in a system that treats every item as interchangeable content. A photograph, rumor, meme, AI reconstruction, restored scan, cropped repost, and researched essay all enter the same feed. The platform flattens the distinctions between them and then expects the audience to determine authenticity in the comments. That is not preservation. It is a system designed for confusion.
A Pew Research Center expert report on the future of online discourse warned that uncivil and manipulative behavior could cause social media to divide into more regulated safe spaces and less controlled free-for-all zones. That prediction feels increasingly accurate. People concerned with standards will not necessarily stop working. They will search for smaller, safer, and more controlled places to put the work. For some, that may mean a personal website, book, private archive, newsletter, moderated community, or curated collection. It may mean publishing fewer images and writing more essays. It may mean using illustrations instead of real photographs. Public platforms may still serve a purpose, but increasingly they will function as signposts rather than homes.
For this reason, I no longer believe Facebook is a safe place for serious photographic sharing. It may remain useful for announcements, links, casual conversation, and material that carries little risk. For serious archival and source-sensitive photographic work, I no longer consider that risk acceptable.
The future may not involve a mass exodus. It may be quieter and more damaging. Valuable images will remain offline, essays will move to personal websites, and larger archival projects will be reserved for more controlled forms of publication. Facebook will no longer be treated as an archive. It will be treated as a place where archives can be stripped for parts. That is what the hollowing out of serious work looks like. It is not the end of Facebook, creativity, research, or preservation. It is the end of trusting Facebook with anything that matters.
Sources Consulted
- Baeza-Yates, Ricardo, and Diego Saez-Trumper. Research on user-generated content and participation patterns on Facebook and Twitter.
- Data & Society. Research on online abuse, harassment, and self-censorship among American internet users.
- Pew Research Center. Reports on online harassment and expert analysis of the future of online discourse.
- Lovato, Juniper, et al. 2024 survey of artists regarding generative AI, consent, ownership, and marketplace effects.
- United States Copyright Office. Report on copyright and generative AI training, including legal and economic questions involving copyrighted works.
- Zhang, Yukun. 2024 paper on generative AI and content platforms, including content oversupply, information overload, and user satisfaction.
