The Unwritten Rules of Grief Tech
AI lets us bring back the dead for advocacy campaigns, courtroom testimony, emotional support, and even entertainment. But is it ethical?
Despite being shot and killed eight years ago, Joaquín Oliver still speaks publicly in support of stricter gun laws. Or at least his AI avatar does.
The high school senior was carrying flowers for his girlfriend when he was fatally shot outside his classroom. He was among the 17 students and staff members killed that Valentine’s Day in the 2018 Parkland school shooting at Marjory Stoneman Douglas High School.
In recent years, Joaquín’s parents use generative AI to digitally revive his voice and image. “I think we’ll see more families choosing to use AI in this way and find themselves making this choice,” said Manuel Oliver, Joaquín’s father.
As AI becomes more convincing and more visible in the public sphere, the ethical frameworks that regulate them are still in their early stages. In the absence of an official rulebook, creators are left to decide for themselves what is acceptable and where the boundaries should be drawn, all while navigating life after losing someone they love.
Grief tech —the emerging industry that uses a deceased person’s digital footprint to construct AI simulations— remains largely self-policed.
Even so, several startups now sell services directly to the bereaved. Grief-tech products range from tools that preserve personal memories to AI companions that simulate face-to-face conversations with the dead. One of them, Seance AI, uses promotional language on its website that suggests AI can preserve the emotional bond between the living and dead.
“Here is where AI meets the afterlife, and love endures beyond the veil,” the company says on its website. “Start your adventure today and let your loved ones speak to your heart once again.”
Grief tech offered Joaquin's father, Manuel, a way to hold his grief and his advocacy in the same place. Through the AI, he could fight for stronger gun laws while keeping Joaquín’s voice present in the conversation.
"I didn't know what I was going to do next," Oliver said, his voice softening as he recalled the day Joaquín was killed. "How do I continue with my life after losing my best friend, my son, my everything?"
Since his resurrection, Joaquín’s AI avatar has been active in the digital advocacy world. He’s been interviewed by journalist Jim Acosta, hosted his own podcast, and spearheaded campaigns such as The Shotline, which uses his voice, and those of seven other gun violence victims to call lawmakers and demand gun reform.
The afterlife advocacy has prompted backlash online, with some viewers accusing Oliver of exploiting his son's death for a political agenda.
"Using the dead like this is really pathetic," one person wrote beneath The Shotline’s YouTube video. Another was harsher. "What disgusting, ghoulish behavior. You're using your dead son to push your anti-freedom, anti-American politics."
Oliver counters that criticism by pointing to Joaquín’s digital footprint as evidence that his son supported stricter gun laws while he was alive.
Joaquín’s once-active Twitter account reads like a record of an ordinary teenager’s interests, filled with Miami Heat posts, animal GIFs and rap lyrics. But buried among them is a post that now carries an unsettling weight. Exactly two months before Joaquín was killed in Parkland, he retweeted a criticism of the NRA following the 2012 Sandy Hook school shooting.
“This is open information,” Oliver said. “You can see it in his posts, his social media and his videos. It made total sense, and again, it was fair to Joaquin to use that direction. So technically, we work with and for Joaquin in this fight.”
As AI becomes more accessible, Oliver said he believes more families will make use of grief tech.
“We were pioneers in doing this, not because we were bringing back the voice of our son, which for some people was a terrible decision, and I totally disagree with. But because we are inviting other families to be a part of it.”
Bruce Green, a legal ethics professor at Fordham University, is concerned that the emotional pull of AI resurrections is reaching the public before people fully understand how it works. Families involved in the process may know the digital likeness was assembled from existing material while the public often encounters only the finished result.
“It’s kind of manipulative, right?” Green said. “It’s a way to play on the sympathies of the person watching it. It has the appearance of being a real person, and you know intellectually it’s not a real person, but you may end up suspending your rational self.”
While no universal standard for disclosing AI-generated videos exists, efforts to create clearer guidance are underway. UNESCO's Recommendation on the Ethics of Artificial Intelligence, for example, states that people should be informed when they are interacting with AI and that creators should be transparent about its use.
Use of AI to bring back the dead was on full display in a courtroom in Arizona in 2025.
In the first known case of an AI-generated victim impact statement presented in a U.S. courtroom, the technology was disclosed right off the bat.
“Hello. Just to be clear for everyone seeing this, I am a version of Chris Pelkey recreated through AI that uses my picture and my voice profile,” the AI ghost says as soon as the video begins.
Pelkey was a 37-year-old Army veteran who was killed in a road-rage shooting in Arizona in 2021. Four years later, his likeness appeared during the sentencing of the man convicted of killing him.
Wearing a grey baseball cap and green hoodie, his AI avatar addressed the defendant directly.
“To Gabriel Horcasitas, the man who shot me, it is a shame we encountered each other that day in those circumstances,” it said. “In another life, we probably could have been friends.”
Pelkey's sister, Stacey Wales, told NPR she co-created the video after watching the trial reduce her brother to crime scene photographs and grim autopsy evidence. She wanted the court to see Pelkey for himself and meet the forgiving, devout Christian behind those images.
Wales drew from family photographs and a YouTube video in which Pelkey discussed God, community and his military service in Afghanistan. She also collected stories from his Army friends and combed through his Facebook posts for speech patterns and lingo.
As the video played, Horcasitas, the man convicted of killing Pelkey, buried his face in his hands and wept. Judge Todd Lang, who presided over the case and later sentenced him to 10 years in prison, praised the presentation.
"I loved that AI, thank you for that. As angry as you are, as justifiably angry as the family is, I heard the forgiveness," Lang said.
Green at Fordham University questions whether the court was moved by the technology itself, or by the fact that it delivered forgiveness. In different hands, he notes, the AI could just as easily have been designed to demand retribution.
“It’d be interesting if the family created a video with the deceased saying, ‘I want an eye for an eye,’ and whether there would be an objection to that,” Green said.
Neither the defense nor the judge objected to Pelkey’s AI presentation in court. Green argues that its admission does not automatically resolve the ethical questions surrounding it.
“If the judge allows it, then you can use whatever techniques the law and judge permit that are effective,” Green said. “I don't know if I were the judge, whether I would accept that form of video because really at the end of the day [in a courtroom], you're either presenting facts or you're presenting arguments. And this is neither.”
While Oliver and Pelkey’s AI resurrections were created directly by their families, a different question emerges when someone’s likeness is recreated without official consent.
Kali-Ma Karefa Smart wasn't aware that an AI chatbot of her grandmother existed until I told her about it. She said she needed a few days to figure out what to make of it.
"There's so much loud conversation about whether AI is good or bad, and less about what it can actually do," Smart said.
The chatbot, hosted by the AI platform DeepAI, was of the late Nobel Prize-winning author Toni Morrison.
Smart said she has no intention of speaking with it herself. “Ever,” she added firmly. Yet she is reluctant to dismiss the project entirely. Its educational purpose gives it a legitimacy she does not see in other AI projects.
"There's at least a mission around it," she said. "Whereas other AI platforms can feel exploitative or useless."
"I believe my Grand would really hope and trust that the curiosity of the user would go far beyond what AI could provide," she said. "That the reader would feel compelled to use their imagination and intermingle with hers."
AI’s educational potential is also what draws Phil Sexton to AI resurrections.
The broadcast news producer traces his appetite for learning to a line he heard as a young Black man: “The best way to hide something from a Black person is to put it in a book.”
The remark became a turning point, convincing Sexton that knowledge could be a form of empowerment. He said he now sees AI as the latest tool in that pursuit.
In his free time, Sexton produces AI-generated short-form vertical videos that place significant Black historical figures, including Martin Luther King Jr. and Frederick Douglass, within fictional conversations about contemporary issues.
In Frederick Douglass Dispatch, an AI-generated Douglass delivers a mock television news broadcast, declaring, “White supremacy did not die, it digitized,” as he discusses the dangers of Project 2025.
Sexton said his AI projects feature only historical figures and fictional characters because recreating living people is “crossing a line that feels tricky."
“For my work, I have a thing for historical moments and characters. So I thought [using] those were okay. As far as contemporary living people, I wouldn't recreate them. I think it's going to be up to the individual user to decide,” Sexton said.
Sexton’s AI videos require a considerable amount of tinkering. His editing background shows in the way he describes his creative process, from animating still frames with Grok and testing voices on ElevenLabs to refining the dialogue in ChatGPT and adjusting the pacing in CapCut. Each video goes through at least three rounds before it arrives at a final cut.
He said he also uses multiple AI models to verify the information in his videos. Sexton asks each model the same question and compares their responses. When the answers align, he takes that agreement as a green light that the information is reliable.
“I try to check claims, historical context, and tone against other sources or AI systems when accuracy matters,” Sexton said.
This verification process reflects a growing tension in AI-generated media, however.
Large language models scrape information off the internet to generate a response, drawing from many of the same source materials to construct their responses. As a result, they can repeat and reinforce the same conclusions, even when that information is incomplete or inaccurate.
This problem extends beyond factual accuracy. AI models can absorb a lifetime of books, speeches and interviews, but cannot capture the nuances of who the person was.
For this reason, Smart argues that AI resurrections have the potential to flatten the dimensionality of human beings into their most recognizable ideas. Toni Morrison becomes the voice of race and memory, Frederick Douglass the voice of abolition, and Martin Luther King Jr. a symbol of equality.
The complexity that made them human—their opinions, private relationships, humor, frustrations and contradictions—is harder to preserve and replicate.
Smart explained that even people who knew her grandmother intimately understood different versions of Toni Morrison but never the full scope.
"People don't know that [Morrison's] closest friend, Eileen Ahearn, never read her books. However, if you spent 20 or 30 years with her, you've read Beloved, you've read Song of Solomon," Smart said.
Some of the earliest signs of AI resurrections came to life on concert stages before thousands of fans.
In 2012, a hyperrealistic hologram of Tupac Shakur appeared alongside Dr. Dre and Snoop Dogg at the Coachella Music Festival. Wearing his signature sagging jeans and Timberland boots, the late rapper shouted “Coachella!” to the 75,000 concertgoers.
Six years later, Prince appeared during the Super Bowl halftime show through archival footage projected across a massive billowing fabric screen. Although the performance technically did not use AI, it reignited questions about consent after fans pointed to Prince's criticism of the technology.
Nowadays, the dead are appearing in far less extraordinary places than sold-out concert stages. Michael Jackson, the late singer and dancer, recently appeared on my social media feed sitting casually among Chicago Cubs fans at a baseball game.
The AI-generated image was part of the “Korean baseball trend,” where users create realistic videos of themselves appearing in the crowd during a live broadcast. Stadium lights, a team jersey, and the grainy texture of television footage recreated Jackson’s 1980s appearance convincingly enough to blend into an ordinary social media feed.
This lack of spectacularity is part of what makes the video significant. AI recreations are becoming more realistic and easier to produce, and images of the dead are increasingly circulating alongside routine forms of entertainment.
In the future, AI resurrections may become personalized experiences in which users can customize interactions, choose when to engage, or even opt out entirely.
Joaquín Oliver’s father said he believes more families will eventually make the same choice.
“Unless there's a really smart reason not to, I think we'll see more families choosing to use A.I. in this way and find themselves making this choice,” he said.
