Pandora’s Chatbox
Updated: Sep 2
Ai isn’t going to take your job; it just might devalue you as an artist and professional. Hop on board the billionaires know best.
Introduction
Ai created content and advertisements have been saturating social media more as the technology evolves. Authentically made art is becoming harder to identify amid image/video generators learning to be more accurate and passable as human art. Although alarming I see a light at the end of this tunnel where art evolves into something AI cannot replicate.
Most painters we think of as the foundation of present day aesthetics, were born in a time when they acted as the sole technology to capture life around them. Landscapes you would never see unless someone painted them. Portraits of historical figures as well as scenes of common life. With artists, literature could be seen from visual representations, such as holy scripture coming to life in enormous paintings. Yet now we live in a world where Donald Trump posts images of Ai art with him representing Jesus Christ.
As art faces the threat of evolving technology, we must remind ourselves that this is nothing new. A new technology arriving and being a tool to create art is a tale old as time. From blowing pigment over hands, to sculpting marble, all the way to creation of cameras and digital art, and everything in between. Especially in the development of digital technologies, art has evolved as a reaction. Cameras developed images in minutes which at the time would take an artist months to replicate. When the world was introduced to cameras, painters embraced a more impressionistic portrayal that captured a humanity cameras could not. Yet these representations were noticeable different; as we face Ai generated content the ability to differentiate human made vs. machine made art is becoming more difficult.
To understand the degree of threat Ai holds towards artists we must research the history of Ai image and video generation. In this report we will investigate how we got to this point, and how society has viewed Ai generated content as it advances. At the end we will have an idea of whether our future is bleak, or on the edge of a second renaissance.
Scientific Foundation: Ai Image Captioning (2015-2017)
In 2015 Ai that could describe the content of images reached a milestone. It was trained with multiple variations and the ability to learn; to correct its mistakes and adapt (Xu et al., 2015). This was truly state of the art technology at the time and in a sense unlocked Pandora’s box. This development opened others eyes to the possibilities ahead, using Ai image labeling as a backbone.
Computer scientist Scott Reed, among others, wanted to flip the process and have captions create images; releasing a research publication stating, “…we developed a simple and effective model for generating images based on detailed visual descriptions. We demonstrated that the model can synthesize many plausible visual interpretations of a given text caption” (Reed et al., 2016).
Many would consider this model as the basis for what we know as image and video generators present day; yet at this point there is still a long way to go. A ‘plausible visual interpretation’ more or less means, if you squint it almost looks like the prompt. As artificial intelligence does, in the following years these image generators would increase in resolution, accuracy, and controllability.
The Public is Starting to Become Aware (2018-2023)
These developments and research were mainly unknown to the masses; until 2018 when for the first time a piece of art generated by artificial intelligence was sold at auction for $432,500. A Paris-based collective named, Obvious, created and auctioned this piece at Christie’s auction house. Obvious member Caselles-Dupré states, “We fed the system with a data set of 15,000 portraits painted between the 14th century to the 20th. The Generator makes a new image based on the set…" (Christie's, 2018). This event ignited debates in the art world and online whether or not what these machines make can be considered actual art. Abstract gallery pieces fought over by mega wealthy for a tax break is one thing, however artificial intelligence was about to threaten a more ordinary artist. With the introduction of Dall-E, and similar technologies in 2021, the debate now shifts from ‘is this real art?’ to ‘could this take my job as an artist?’.
On January 5, 2021 company OpenAi demos a text to image model named DALL-E while also releasing a program called CLIP which turned images into text as a means to identify what DALL-E generations were accurate representation of the prompt (Johnson, 2021). Although DALL-E was not released to the public, the accessibility of CLIP allowed independent developers to reverse engineer the technology and create a text to image generator of their own. Most notably was the technique implemented by Ryan Murdock called Big Sleep which paired CLIP with existing deep learning architecture referred to as GANs (Generative Adversarial Network). Computer scientists at the EvoStar Conference in 2021 reviewed this method, "We have found that CLIP-guided GAN text-to-image generation produces images which are routinely appropriate to the prompt, usually have artistic value, are often innovative internally and have high diversity across a population for the same prompt. Moreover, the Colab notebooks are highly accessible as they require no technical background, and hence could rapidly influence mainstream art production, given the evidence of valuable usage presented here” (Smith and Colton, 2021).
As images generated with this technique begin to go viral on social media the technology continues to improve in the background. Big Sleep acquires a new CLIP method with VQGAN being implemented which improves accuracy. At the same time David Holz recruits ten engineers to train diffusion models resulting in the first private demo of Midjourney in September 2021 (Kayman and Segal, 2025). Both text to image models from Midjourney and OpenAi continue to be improved and in 2022 the public starts to gain access.
To begin 2022 different versions of DALL-E and Midjourney were released in beta forms open to small private Discord groups only. That is until July 12th when Midjourney opened their usable Discord bot to the public and grew to a million users when V3 was released two weeks later (Ma, 2023). Possibly because of the ongoing momentum of a competitor, DALL-E removed its waitlist in September and right away had, “over 1.5M users creating more than 2M images per day” (OpenAI, 2022). The flood gates have opened, what will come next?
Lawsuits are Filed and Text to Video is Born (2023)
Backed by years of research and development, the massive public usage dramatically improved text to image generation. These results began to create debates on the issue of copyright, and the lawsuits began. Absorbing masses of content with internet scrubbing, by order of developers, the models have valuable image resources to reference. With assumably billions of reference photos AI could now even replicate the styles of human artists in a believable way. Artists, Sarah Anderson, Kelly McKeran, and Karla Ortiz saw this development as a threat to all artists and filed a lawsuit against Stability AI (creator of Stable Diffusion), Midjourney (creator of Midjourney), and DeviantArt (creator of DreamUp) for copyright infringement, DMCA violations, and related state law claims in January 2023. The basis of their lawsuit can be understood in the following art law synopsis, “the plaintiffs argue, instead of commissioning an (artist) to create a work, or paying for a print or license of the work, now users simply prompt an AI image generator to create an artwork in the style of a specified artist – one that is ‘indistinguishable from one the artist might’ve created themselves’” (Williams, 2024).
Another notable lawsuit was filed on February 3, 2023 by Getty Images which accused Stability Ai of using more than 12 million of their licensed photos without permission to train their models. The complaint claims, “Stable Diffusion at times produces images that are highly similar to and derivative of the Getty Images proprietary content that Stability AI copied extensively in the course of training the model” (Setty, 2023). These cases illustrate the continuing legal uncertainty surrounding copyright, training data, and generative AI as both cases continue litigation to this day because of the nuance in the uncharted territory of Ai copyright. The world of artists is shaken, but the biggest threat is yet to come.
Research from 2022 describes using existing text to image models to create a new technology, “it is natural to consider leveraging image priors for videos to simplifying the learning process. After all, an image is a video with a single frame. In unconditional video generation, MoCoGAN-HD formulates video generation as the task of finding a trajectory in the latent space of a pre-trained and fixed image generation model” (Singer, et al., 2023). In late March 2023 text to video generation graduated from research demonstrations to a usable tool; and within days of each other, text to video models ModelScope and Runway were publicly available.
While image generators at this point could be photorealistic and fixed major issues such as depicting hands, video generation was far from human. Aaron Mok reports on how the public was starting to use ModelScope Text to Video Synthesis, “There's a nightmarish compilation of ‘Will Smith eating spaghetti’ that has racked up 1.8 million views on Twitter and depicts a cartoon-like version of the … star shoving handfuls of spaghetti into his mouth. … Viewers have called the videos ‘cursed,’ ‘deeply unnerving,’ and ‘super creepy’” (Mok, 2023). Humorously, the public would use this prompt of Will Smith eating spaghetti as an ongoing test for the performance of text to video Ai models. Video generation is now accessible and people other than artists begin to be startled, especially as machine generated images are becoming more believable as reality.
Misinformation Runs Rampant (2023-2024)
“Extrapolating to the political context, the pervasiveness of meta-cognitive myopia suggests that citizens are particularly vulnerable to misinformation… the environment of online social networks which provides users with easily accessible meta-data like the number of likes and shares of a given piece of (mis)information is likely to be myopically interpreted” (Pantazi et al., 2021). Decades of psychological studies show our bias-information processing and other cognitive bias leave humans susceptible to gullibility especially in a political context. These mental functions have been dramatized by social media with the mass quantity of repetition and the number of likes and shares, which influence viewers to perceive that information as factual. Now with Ai image generators becoming hyper realistic we can only assume this line distinguishing real and fake information will become blurrier leaving many susceptible to misinformation.
In March of 2023 Donald Trump was photographed entering Manhattan criminal court to face a 34-count indictment, these pictures began to spread on social media along with Ai generated images of Trump. “Some of the images, which were fabricated, appeared to be mug shots of the former president, even though his lawyers told reporters the former president did not take a booking photo during police processing on Tuesday” (Saliba, 2023). Although these mugshots were mainly known to the public as fake, they were used as a marketing tool in Trump’s successful 2024 presidential campaign. One of Trump’s first of many notable usages of Ai creations for his political messaging.
An example of Ai successfully tricking the masses can be found the same month, when Pope Francis was receiving fashion praise for a trendy puffer jacket. “The AI-generated images of Francis… widely circulated on Twitter. The images fooled scores of users, in one of the first instances of wide-scale misinformation stemming from artificial intelligence” (Tolentino, 2023). This phenomenon caused concern as a joke created in Midjourney was able to spread believable misinformation. The growing worry surrounding evolving Ai powered tools is validated when sexually explicit deepfakes target someone who may not be a world leader, but is just as influential to certain demographics.
In late January 2024 it is reported on NBC news, “Nonconsensual sexually explicit deepfakes of Taylor Swift went viral on X… amassing over 27 million views and more than 260,000 likes in 19 hours before the account that posted the images was suspended.” The article goes on to state, “Such images can be generated with AI tools that develop entirely new, fake images, or they can be created by taking a real image and ‘undressing’ it with AI tools” (Tenbarge, 2024). The dark side of Ai image generation is presented to the world with other sources estimating the number of views on these pornographic images to be between 40-50 million.
In direct response to this event U.S. Senate Majority Whip, Dick Durbin, along with other senators introduced the, Disrupt Explicit Forged Images and Non-Consensual Edits Act of 2024 (DEFIANCE Act). This would criminalize the spread of nonconsensual sexualized Ai generated images. Durbin states, “This month, fake, sexually-explicit images of Taylor Swift that were generated by artificial intelligence swept across social media platforms. Although the imagery may be fake, the harm to the victims from the distribution of sexually-explicit ‘deepfakes’ is very real” (Montgomery, 2024). As seen with this example, Ai generated content is now accurate enough at depicting reality that it requires ongoing legal conversations and actions.
Ai Begins to Scare Hollywood (2025-2026)
In 2025 Ai image and video generators are publicly available with access to many programs being free. Hollywood has already been on high alert for years at this point, as is apparent from the 2023 writers’ strike focusing mainly on Ai text generation. The topic of Ai deepfakes of actors was also a point of contention, however this only resulted in an agreement to follow terms for consent and compensation (Patten,2023). It was not until September 2025 when OpenAI launched Sora 2 as a standalone app that this issue needed more attention.
OpenAi’s text to video model, named Sora, was available to the public in December 2024 with the second version being released in app form to the U.S. less than a year later. The public were able to generate videos replicating prominent figures in an arguably believable way. What may have begun as jokes with Ronald McDonald quickly shifted to representing deceased public figures such as Martin Luthur King Jr., Kobe Bryant, Tupac Shakur, among others. Popular living actor Bran Cranston was also among those represented stating, “I was deeply concerned not just for myself, but for all performers whose work and identity can be misused in this way” (Campione, 2025). At times the only way to differentiate between reality was the Sora watermark and the absurdity of the situation.
CEO of OpenAI Sam Altman commented on the situation, “OpenAI is deeply committed to protecting performers from the misappropriation of their voice and likeness… We were an early supporter of the NO FAKES Act when it was introduced last year, and will always stand behind the rights of performers” (Yip, 2025). The NO FAKES Act is a proposed bill which began to be discussed in late 2023 with the sole purpose to, “protect intellectual property rights in the voice and visual likeness of individuals, and for other purposes” (S.1367, 2025). To this day the bill is still yet to be passed by congress.
Even with OpenAi supporting this act and putting guardrails in place in response to backlash, the damage was done. Sora 2’s introduction to the masses generated content that was not just seen as a joke anymore. A less tech savvy generation now becomes confused on what is real or fake on the internet, leaving them confused and susceptible to a rising amount of Ai product scams. At a U.S. Senate Special Committee on Aging hearing, chairman Rick Scott stated, “In 2025, Americans over 60 lost a staggering $7.7 BILLION to scams. And that’s just what scams have been reported. The rise of Artificial Intelligence, AI, has offered scammers new tools to pursue their criminal schemes and we must adapt and respond to these new threats” (Scott, 2026). He goes on to explain that products are advertised with deep fakes of trusted experts or celebrities, and Ai generated voices of loved ones are reported to be used in scams in what Scott describes to be, “heartbreaking and evil ways.”
A simple comparison of a ‘will smith eating spaghetti’ prompt on ModelScope vs. Sora 2 would make a Hollywood executive frightened. With a now complete audio and visual generation, the threat to the film industry is no longer hypothetical, it is here and exponentially growing every moment. Sora 2 shocked the world but the public attention it was receiving did not reflect in active users, especially after more copyright restrictions were introduced. The app was costing OpenAi money to host with little return; and surprisingly even after Disney pledged to invest a billion dollars into the company, Sora 2 was shut down March 24, 2026 (Masunaga, 2026). Without a statement from OpenAI explaining its decision, questions remain as to whether public backlash, financial costs, and ongoing legal conflicts played a role in the company stepping back from video generation and refocusing its efforts on their LLM ChatGPT which contains image generation.
Sora 2 left no void with its departure as other companies continued to improve their Ai text to video models. Released in July 2026, Seedance 2.5 and Minimax’s Hailuo 3 show us content that is quite often hard to differentiate between human made content. Music videos, advertisements, full storyline Disney type animations; with the right script and modifications all look so real the companies are receiving cease and desists (Kanter, 2026). Copyright battles are ongoing, but beyond IP owners, many creatives watch nervously wondering where this technology will evolve from here.
Reflection
At the end of the day Ai art is just another subculture of art. As needles are to knitting, copywrite is to ai art generation. It is the art of a guiding a computer’s hand so to speak; it can be that and your hand painted masterpiece can be something even greater. That being said, it is apparent through the timeline of these technologies that we must proceed with caution. The main issue does not seem to be the Ai models themselves, rather their ability to produce content that infringes on others hard work.
Either out of financial and political concerns or ignorance, the government does not seem to carry much urgency in this matter. These technologies will only evolve and I hope to see more laws be passed to protect not only artists but the masses from something that can be used immorally.
Putting the issue of copyright aside, I believe as time goes on the general public will be able to tell the difference of time, emotion, and effort put into human vs. machine made art. Ai generated images and videos have telling signs it was not made by a human in many cases; although referencing existing work, it does portray as its own style/genre. As it learns to fix its imperfections we humans will also get better at identifying those generations.
As an artist I am not upset this technology exists, in fact it motivates me. I strive to create pieces no one has seen anything like before, something Ai could not replicate. I strongly believe that Ai art will inevitably create higher demand for true human art. In reaction to this technology, a second renaissance could be on its way that will change art as we know it forever.
Works Cited
Allyn, Bobby. "OpenAI Blocks MLK Jr. Videos on Sora after 'Disrespectful Depictions.'" NPR, 17 Oct. 2025. Houston
Public Media, www.houstonpublicmedia.org/npr/2025/10/17/nx-s1-5577869/openai-blocks-mlk-jr-videos-on-sora-
Campione, Katie. "Bryan Cranston, UTA, CAA & SAG-AFTRA Praise OpenAI's New 'Guardrails' for Sora 2 After
'Breaking Bad' Actor's Image Was Used Without Consent." Deadline, 20 Oct. 2025,
Christie's. "Is Artificial Intelligence Set to Become Art's Next Medium?" Christie's, 12 Dec. 2018,
Johnson, Khari. "OpenAI Debuts DALL-E for Generating Images from Text." VentureBeat, 5 Jan. 2021,
Kanter, Jake. "Chinese AI Is Coming for Hollywood: ByteDance, MiniMax and The Office." Deadline, Aug. 2026,
Kayman, Alexa, and Etienne Segal. "Midjourney." Contrary Research, Contrary LLC, 23 May 2025,
Ma, Kev. "All Midjourney Versions (V1-V6) Compared: The Evolution of Midjourney." AiTuts, updated 20 Mar. 2023,
Masunaga, Samantha, and Nilesh Christopher. "OpenAI Will Shut Down Its Sora Tool." Los Angeles Times, 24 Mar.
2026, www.latimes.com/entertainment-arts/business/story/2026-03-24/openai-will-shut-down-sora-why-what-to-
know.
Mok, Aaron. "I Can't Stop Watching These Hilariously Bad AI-Generated Videos of Celebrities like Will Smith and
Scarlett Johansson." Business Insider, 29 Mar. 2023, www.businessinsider.com/watch-hilariously-bad-ai-
Montgomery, Blake. "Taylor Swift AI Images Prompt US Bill to Tackle Nonconsensual, Sexual Deepfakes." The
Guardian, 30 Jan. 2024, www.theguardian.com/technology/2024/jan/30/taylor-swift-ai-deepfake-
nonconsensual-sexual-images-bill.
OpenAI. "DALL·E Now Available Without Waitlist." OpenAI, 28 Sept. 2022, www.openai.com/index/dall-e-now-
Pantazi, M., et al. "Social and Cognitive Aspects of the Vulnerability to Political Misinformation." Political Psychology,
International Society of Political Psychology, 2021, www.onlinelibrary.wiley.com/doi/10.1111/pops.12797.
Patten, Dominic. "Read Full SAG-AFTRA Deal Summary Released Ahead of Ratification Vote; Analysis of AI &
Streaming Fund." Deadline, 12 Nov. 2023, www.deadline.com/2023/11/sag-aftra-deal-summary-released-read-it-
Reed, Scott, et al. "Generative Adversarial Text to Image Synthesis." Proceedings of the 33rd International Conference
on Machine Learning, 2016, pp. 1060–69, www.arxiv.org/abs/1605.05396.
"S.1367 - NO FAKES Act of 2025." Congress.gov, Library of Congress, 9 Apr. 2025, www.congress.gov/bill/119th-
Saliba, Emmanuelle. "'A Real Worry': How AI Is Making It Harder to Spot Fake Images." ABC News, 7 Apr. 2023,
Scott, Rick. Chairman's Opening Remarks: July 29th Hearing. U.S. Senate Special Committee on Aging, 29 July 2026,
Setty, Riddhi. "Getty Images Sues Stability AI Over Art Generator IP Violations." Bloomberg Law, 6 Feb. 2023,
Singer, Uriel, et al. "Make-A-Video: Text-to-Video Generation without Text-Video Data." International Conference on
Learning Representations, 2023. arXiv, 29 Sept. 2022, www.arxiv.org/abs/2209.14792.
Smith, Amy, and Simon Colton. "CLIP-Guided GAN Image Generation: An Artistic Exploration." Evo* 2021 — Late-
Breaking Abstracts Volume, arXiv, 22 June 2021, www.arxiv.org/pdf/2106.11804.
Tenbarge, Kat. "Nude Deepfakes of Taylor Swift Went Viral on X, Evading Moderation and Sparking Outrage." NBC
News, 26 Jan. 2024, www.nbcnews.com/tech/misinformation/taylor-swift-nude-deepfake-goes-viral-x-platform-
Tolentino, Daysia. "AI-Generated Images of Pope Francis in Puffer Jacket Fool the Internet." NBC News, 27 Mar. 2023,
Williams, Sophia. "AI and Artists' IP: Exploring Copyright Infringement Allegations in Andersen v. Stability AI Ltd."
Center for Art Law, 26 Feb. 2024, www.itsartlaw.org/art-law/artificial-intelligence-and-artists-intellectual-property-
Xu, Kelvin, et al. "Show, Attend and Tell: Neural Image Caption Generation with Visual Attention." Proceedings of the
32nd International Conference on Machine Learning, 2015, www.arxiv.org/abs/1502.03044.
Yip, Jaures. "OpenAI Cracks Down on Sora 2 Deepfakes After Pressure from Bryan Cranston, SAG-AFTRA." CNBC, 20
Comments