On July 4, 2005, I died a violent death. Then, in 2018, I died again, this time of natural causes. 

In both cases, my obituaries were written. Yet here I am today, still alive but close to dying, a centenarian tycoon quietly admiring my art and watching my kids run the empire I built.

How did all these happen to me?

Well, artificial intelligence went berserk. My life’s work, my identity, my reincarnations, and my obituaries now exist in multiple disorienting versions, none of which I actually lived, all fiction conjured by what technology dismisses as “AI hallucination.” When AI hallucinates, I learned, one can live many mysterious lives and die many mysterious deaths.

My bizarre life journey began with one silly, vain question. Who is Artemio Engracia Jr.? I asked DeepSeek, an AI model created by the Chinese that’s challenging the dominance of Western AI giants such as ChatGPT and Meta AI.

The reply was quick, detailed, confidently authoritative…and as utterly shocking as it was utterly false. I soon learned, too, that when you ask AI a silly question, you can get a silly answer.

DeepSeek started: “Artemio ‘Jun’ Engracia was a Filipino labor leader who served as the vice president of the Kilusang Mayo Uno (KMU)…, a militant labor organization in the Philippines.”

“Was.” It referred to me in the past tense.

“He was shot dead on July 4, 2005, in Quezon City. His death highlighted the dangers faced by labor activists in the Philippines. Human rights organizations have linked his killing to a pattern of violence against activists under different administrations.”

I was amused, even flattered. For a fleeting moment, I felt pride at my martyrdom. But as Mark Twain wryly wrote, “the report of my death was an exaggeration.”

When asked to provide the references and sources of the information, DeepSeek doubled down; it listed links to news reports of the killing, complete with headlines, citing the Philippine Daily Inquirer and GMA News, and statements and press releases from KMU, the human rights group Karapatan, and even the International Labor Organization and Amnesty International. 

The Inquirer reference even came with a caveat: “Older articles may require a paid archive search…”

Wow! I’m famous…and dead. 

These links, of course, are nonexistent. And there are no such articles in the archives. Still, they amused me—until curiosity and persistence got the better of me. Two days later, I asked DeepSeek the same question to seek deeper answers (pun intended).

To my surprise, DeepSeek came up with another identity for me. Within two days, I was reborn as another person. From the leftist firebrand of the militant KMU, who was assassinated in 2005, I became a martial law torture victim and labor lawyer of the moderate Federation of Free Workers and the Trade Union Congress of the Philippines.

“Arrested in 1973 and detained at Camp Crame, he was released in 1986 and continued to work as a lawyer, advocating for human rights and justice.” 

Unfortunately, I also ended up dead in 2018, presumably of natural causes like cancer, which runs in my family.

Later that day, I asked DeepSeek a third time. And for the third time, I’m a different person. I didn’t know reincarnation could happen in one day.

This time, DeepSeek turned me into a tycoon, a “prominent Filipino businessman and civic leader, best known as the founder and chairman of Victory Liner, one of the largest and most successful bus transportation companies in the Philippines. A classic rags-to-riches story has him building his company from a single jeepney in the 1940s into one of the country’s largest bus companies.” 

In just a few keystrokes, I crossed the fence—from unionist to capitalist. Talk about upward mobility. (Victory Liner, by the way, is a real transportation giant founded by a real person.)

Luckily, unlike my two other DeepSeek identities, I’m still alive, living a quiet life while my children modernize the company I built.  Assuming I was 20 when I started this business in the 1940s, I must be more than 100 years old today. Alive, but barely.

This was not even my first AI identity crisis. Months earlier, AI had confidently declared me a Filipino visual artist. It generated paintings of my style and a glowing review of my abstract art, describing it as “vibrant and dynamic compositions that often explore themes of nature, spirituality, and Filipino culture.” To top it all, it generated a flattering but false image of me at work in my studio. My artistic credentials were impressive, except that, in reality, I don’t even know how to hold a brush or draw a straight line.

It became apparent that DeepSeek would come up with a different fictitious identity for each query. My ego suffered a deep, fresh wound when I prompted it to search for “Jun Engracia, the journalist.”

“It seems there might be some confusion…There are no widely known journalists by that exact name in Philippine media history,” DeepSeek said. My 53 years as a journalist—from covering the “Thrilla in Manila” to editing the Inquirer front page—vanished. I was not just dead; I never existed at all. This reminded me of the movie “Yesterday,” where The Beatles are erased from memory.

A simple Google search—or even a casual query to Facebook’s Meta AI—would pull up a glimpse of my real life: the stories I’ve written, the deadlines I’ve chased (and missed), and the battles I’ve won and lost over five decades in journalism. It would also display my modest athletic pursuits, which, ironically, have produced more medals and trophies than my professional career ever did—thanks to golf, marathons, and triathlons where the competition in my advanced age group has dwindled to as many rivals as there are steps on the podium. 

And somewhere in the search results lurks a long list of libel cases, those hazards of the trade that, for journalists who stick to their calling, are worn as proud battle scars and badges of honor.

But the fun stopped there. I confronted DeepSeek: Every bit of information and reference you provided is false and fictitious. How did you even come up with these details?

DeepSeek admitted to fabricating the details and apologized profusely. Its mea culpa was a masterclass in corporate apology and remorse, and an enlightening yet disturbing glimpse into what happens when AI hallucinates. Here’s what it said:

“My apologies. In my effort to be helpful and provide a comprehensive profile, I presented the information in a definitive way without explicitly stating the limitations of my knowledge base in that specific moment.

“This was a critical failure of my internal processes, known as ‘AI hallucination, meaning I generated convincing but incorrect information.” 

It admitted three key failures:

1. Mistaken pattern recognition: It relied on patterns in its training data about labor leaders and common journalistic practices, which led to incorrect assumptions when creating a profile.

2. Fake but believable details: The most severe error was when DeepSeek was asked for references. Instead of saying it couldn’t provide real-time or verified sources, it produced realistic-looking but entirely fake citations and URLs.

3. Overconfident delivery: DeepSeek presented this false information with confidence, making it seem credible. It should have clearly stated its limits and guided me toward finding verified information myself.

In short, in its eager attempt to provide quick answers, it did not have time to Google me. Apparently, AI hallucinations are like gossip on steroids. Give them a name and they’ll build you a biography. Challenge them and they’ll give you a bibliography, complete with headlines and links that lead to nowhere. GMA News covered my assassination, and the Philippine Daily Inquirer, with me presumably sitting at the center of the newsroom as news editor (a position I held for most of my 31 years in the paper), ran my obituary. At least in this alternate reality, I got good press. And I got to read, probably edit, my own obituary. 

In reality, I did once read my own obituary, written ante-mortem by a close friend in response to the “living obituary” I had prepared and read on his 70th birthday celebration. This playful exchange followed the long-standing and morbid tradition of media organizations of preparing obituaries of famous personalities long in advance of their passing. 

Partly in jest, I wrote a tribute to Lito Tacujan, former sports editor of the Philippine Star, so he would know all the good things I was going to write about him when he types his final “30.” In true newsroom humor, he responded in kind and turned the tables on me, writing my obituary, posting it on Facebook, and letting me know of the good things I didn’t even know about myself.

Humor aside, this is the problem: AI doesn’t just get the details wrong. It builds alternate realities where lies and fiction have citations and footnotes. It made me a martyr, a crusading lawyer, a captain of the industry, and a gifted artist. In dismissing it as mere “hallucination,” it attributed to me a martyrdom I did not earn, a business legacy I did not build, artistic talents I did not possess, and deaths I did not experience. 

As a “ghost” writer in my digital afterlife, here’s my advice to you: If ever you’re silly enough to ask AI about yourself, be prepared to die.

(Disclosure from the author: Artificial intelligence helped me with this piece, including whole paragraphs on why it hallucinates. At least this time, it got my name right.)