Post

When I left the Washington Post in 2023, my colleagues and I were worried AI would hoover up our work and enable "zombies" to churn out stories based on our reporting and writing styles. Today WaPo announced a partnership with OpenAI. And here it is: Zombie Gillian Brockell. I DID NOT WRITE THIS.

4.1K reposts10K likes

When I took the history reporter job, I told my editor I would never "both sides" the Civil War, Confederates or slavery. I never did. But here, ChatGPT (pic 1) has turned my work (pics 2,3) into whitewashed pablum and put MY FUCKING NAME on it. I am livid, I am shaking. I don't know what to do.

THE POINT OF MY LIFE'S WORK IS THAT LOST CAUSE "HISTORY" IS MYTHOLOGY. It isn't a difference of opinion, it is literally not true. And I have always written like a salty bitch. My editor would ask me to turn up (or sometimes down lol) the salty bitch. Where the fuck is the salty bitch?????

Post an article online about how ChatGP stole your work and then misinterpreted it. Use lots of search words that would link it to this piece so it will then start saying “we lied in this article and stole the original authors work”.

Genuinely consider suing, or at least consult with someone about it. Slam dunk defamation and I bet you could find lawyers who’d be champing at the bit to take it.

This is offensive on a deeply human level and everyone responsible should be ashamed. They won't be, since they are obviously soulless lizard people, but they should be.

a) Sue b) Sue c) Find the master plug for OpenAI and pull it, then (wearing safety gloves) jam a fork in the outlet and short circuit the whole building d) Raze the building to the ground e) Plant trees over its remains f) Sue

Sorry, but I don’t follow. Was it you or someone else who asked the AI to produce this? If not you then who? And if you then are you more upset that it was able to produce anything at all or that it didn’t do a better job?

Just so I’m clear: *you* entered the prompt into ChatGPT & generated the article? If so, it didn’t exist until you asked it to be born. *You* are the one who put your name on it, not AI. Can WaPo do it? Sure. But thousands of writers have been ringing alarm bells about this for a very long time.

Demand that your name be removed from this, insist on a written apology for having your work AI-fucked into your named (not) “voice” that isn’t buried in the last page of the paper, and then get an attorney and go after those twatwaffles, Gillian. Fuck them and the phallic rocket they rode in on.

It's difficult to tell. The UI looks similar. When I tried it (apologies to Gillian Brockell). It's possible it's down to the version, but my version uses the suffix *"-style"*. It's possible this is a screengrab from the WP optimised LLM.

Under U.S. copyright law, work created by an employee within the scope of their employment is considered “work made for hire.” This means: • The employer (the newspaper) is the legal author and copyright holder. • The reporter has no ownership of the material they write as part of their job. (1/2)

You could sue, but you entered into a transaction with WaPo and were compensated, so there’s no question of ownership. Unless you had a “no derivative works” clause (unlikely), then you have no standing for a case against WaPo. You could try to argue impersonation or slander against OpenAI, but GL

Okay but you asked ChatGPT for this output. You created that piece. You published it online. This isn’t to strawman you, but it’s a really important distinction because HUMANS have to be responsible for AI. If someone did this to you, that would be a moment to feel livid. That didn’t happen.

You need an IP attorney who can look at this from an identity theft perspective where your value as a respected writer is being damaged. a combination of false attribution & defamation. It would certainly be cutting new ground. In DC, you might contact Volunteer Lawyers for the Arts for a referral.

Also, you sue everyone and everything, such as WaPo, Bezos, and anything direct or tangential to the AI. Also, call the Authors Guild in NY, which is working against this stuff. If I were to find my late husband’s name used like this, I think his celebrity registration under CA law might help me.

What happens if you prompt something similar without a request for your style? Thinking that if they pull archive prompts what is the risk that they have your material in there? This is a risk for the journalists as well, they don’t want to plagiarise.

The people telling you to sue seem to be confused by the way you framed this. They think it's a direct result of the WaPo deal or something that was actually published by WaPo. This is something ChatGPT could already do, that you asked it to do.

Not a lawyer myself, but I think you may want to talk to a few lawyers. Your local and state bar associations should have resources to connect you with attorneys, but usually you need to know what kind of lawyer you'd need. In lieu of any further insights, I'd personally start with a labor lawyer.

Not a lawyer. Consult one. But there are two angles here: 1) Defamation by OpenAI. That has some legal grounding/similar cases already. 2) WaPo's deal with OpenAI. Note, as others are pointing out, this output is independent of said deal, but the deal is way more concerning. What matters there...

...is what your contract said they can do with your work. So yeah, consult a lawyer about that, but even if this is a "legal" usage, thank you for speaking up and writers need to ban together to MAKE it an illegal usage (i.e., cannot sell work for AI training by default, without explicit permission)

WaPo made a deal with ChatGPT, so it is hard for it to then say it has nothing to do with what ChatGPT is doing with its work, and if WaPo itself posted this article you could very obviously sue, so whatever attorney you would get for that, I'd get for this. And a consultation is probably free.

Misrepresentation, diminishing value of your work, damage to reputation. Good lawyer will pull the list. By adding your byline, this hasn't created a lookalike, but a weak forgery. Like if an art student whipped up something in the style of Klimt and then told people it was an authentic work.

Eventually, as search engines get fed by the same AI puking up content, nothing on the internet will be reliable. Words and images made by dumb machines will be signed and scattered among the legitimate works, and then the fake stuff will bury it.

I think it falls under the umbrella similar to defamation - it is invasion of privacy (misappropriation or appropriation and/or false light). A first amendment lawyer should be a good option or a general practice tort litigator.

Or, imagine going to a WaPo employee and asking him what you asked ChatGPT. Will you sue him because he tried to - badly - imitate you, at you own request, on the argument that the WaPo management could ask him to do the same?

Mmm, it is akin to a painter, let's say XX, going to a counterfeiter and commission him to make a painting in his (XX) style and sign it as XX and give it only to him (XX). I do not see how XX could sue the counterfeiter? The ability to potentially commit a forgery is not a crime.

The Authors' Guild has been organizing a class-action suit against OpenAI and ChatGPT for unauthorized use—you didn't agree to have your work train generative AI. It's one thing for WaPo to use/reprint/repackage your work, but not this. This would be under copyright law/intellectual property.

Don’t sue, but request an opportunity to publish a response. Then, if denied, take it to another outlet until someone runs with it. Then WaPo gets to defend its actions…and look like the fraud they are.

“When to Consult a Copyright Attorney vs. a Human Rights Attorney: Copyright Disputes: If you are facing a copyright infringement claim or have a dispute over the ownership or use of your work, you should consult with a copyright attorney.”

You could have grounds to sue the developers of chatgpt for defamation too. They made a piece of software which claims you did something horrendous that you did not. How amenable a judge would be to that suit, relevant precedent, etc an attorney would have to say

I think people are misunderstanding your screenshot to be something published on WaPo with your name in the byline, rather than something you fear will be published on WaPo in the future.

Defamation. There are putting a position you do NOT hold in your mouth. Spreading lies that affect the public perception of your values. Sue WaPo. They published falsehoods under you name.

Most likely an attorney who knows publicity rights. “Personality rights, sometimes referred to as the right of publicity, are rights for an individual to control the commercial use of their identity, such as NAME, image, likeness, or other unequivocal identifiers.”

You are going to need a lot of help. I would contact the ACLU and get started with rights you have to your own work to not be misrepresented. Go after ChatGPT. Honestly myself and tons of developers are losing work to them and they are giddy about people losing jobs now to use their shitty AI.

That's good context, but it could only happen if WAPO agreed to cough up all your work, and they did. Meanwhile NYT is suing to prevent it from happening to their reporters.

I would look to sue ChatGPT. They have misrepresented your work and damaged your reputation. They should not be making their tool available for use until they can guarantee not to do such things to people's work.

That clarifies things GREATLY. If WaPo had published that? Then yes, sue the f*** out of Bezos. As it's on Chat GPT? I'd consult a lawyer as to who to sue. At the least, it's a poor impersonation which could damage your reputation.

This is the horror of it, I'm sorry it happened but it's also happening to every writer, artist and musician out there - this slop generator steals truth and creativity and regurgitates it to an idiot public. Not good!

MEMORY: Abusive, slaveholding labor thieves willing to kill their own families to continue to hold humans in bondage REALITY: Abusive, slaveholding labor thieves willing to kill their own families to continue to hold humans in bondage I'm not seeing the alleged disconnect.

Sue ChatGPT/OpenAI for defamation and theft of publicity. They are using your name to implicitly endorse a product, their chatbot. They are implicitly saying you would both-sides the civil war and thus wrecking your professional name.

WaPo owns the words I wrote for them, and they have the right to create derivative works, like a movie, from those words, though iirc they had to give the reporter a cut of the deal. They do no the right to use my name OR TO CONFER THE USE OF MY NAME TO A THIRD PARTY.

IANAL but I have a degree/experience in criminal justice This is theft of intellectual property, pure & simple Easy grounds for a lawsuit You are entitled to “own” your own ideas, writing, etc. Having AI emulate you is plagiarism w/ extra steps - but it is definitely theft of intellectual property

Get together with every journalist, voice actor, screenwriter, songwriter, author, etc that has been, or can be possibly be used to “feed” AI and try to destroy AI in court. Make no mistake. The Oligarchy frothingly wants AI to replace everyone because employers hate to pay employees

I’m so sorry this is happening to you. This is so horrible, and I don’t think there are laws yet to cover it. There must be something that can be done. If they’d kept going with your by line and had a human write it would you have recourse? Seems the same thing.

Remove the AI component. 1. Then it's forgery. Whoever authorized this (ultimately Wa-Po) if libel for forgery. 2. Regarding style and opinion they are also liable for infringing on your body of work. Copyright infringement. This is the type of BS id expect from Bezos

Do you know if this prompt worked prior to the deal? I suspect it probably did though now subsequent models will have direct access to the WaPo articles

She asked ChatGPT to write something in her voice. It did. Nobody else did anything. She picked the topic. She is confused because you could ask it to write in anyone's voice and if it knows that person, real or made up, it would try. It might be a problem, just not the one she thinks it is.

Seems like they've pulled it. But wapo has other articles under your byline since 2023. "Complicated legacy" is a firm favourite of wapos headline writer. Dozens of bio articles with that description.

I just want to note that you can give ChatGPT a prompt with literally any name -- real names, fake names, silly names, serious names -- and it will do the exact same thing. Here's an excerpt in the style of extremely not-real WaPo reporter Barnabas Flimflamington. This outrage over this is silly.

ah yes the very recent development of people suddenly thinking maybe Robert E Lee was not a good guy fighting for a cause that's not worth naming but definitely tragic in its failure.

Shame on WAPO for the partnership. Shame on society for letting this type of thing exist. Shame on OpenAI for not tuning it to not be able to say "by" even if requested. But the fact this is legal is the entire problem. People claiming it isn't are missing the greater horror.

Most of the comments about what this legally is are completely incorrect. This isn't fraud, it isn't a copyright violation, etc. It's unethical. It's gross. But there doesn't seem to be anything illegal in the response to this _prompt_ asking to write something "in the voice" of another.

I’m really sorry this is happening to you. This is shit. I hope you’re able to protect your mental health and reputation through it all. Total shit.

Are we at the end of written media? If the Washington Post is going to be written by robots, what is the point of humans reading it? If we can’t tell what’s written by robots anymore, is anything written after today worth reading?

But… ChatGPT could do this yesterday. Or a year ago. I guess you can be mad that WaPo is partnering with any AI at all, but this isn’t some malicious intent by anyone to pass you off as a Confederacy supporter. You’re the one who asked ChatGPT to do this.

Absolutely, but my point is that it wasn’t caused by WaPo’s partnership. ChatGPT could be told to generate a particular position in a particular voice already. Your issue is with all AI, not this most recent announcement.

My issue is with both. Did WaPo have the right to enter into an agreement with OpenAI if this harm was part of the outcome? This feels very much like externalized costs to me; as with environmental impacts from business, the human costs of AI are being elided.

This is the red line, the end of my WaPo subscription. Is there any national newspaper worth subscribing to anymore. One that doesn't use AI and bend the knee to the fascists and try to bust its employees unionization efforts? Or am I going to have to get all my US news from overseas?

I’m sorry your name is on it! I was a product description writer and lost my job just over a year ago. I’d had no idea that small businesses were already using ChatGPT so ubiquitously. I’ve noticed all the ChatGPT ads use the word “Elevate.” That word makes me cringe now.

Hey Gillian, just to be clear - this is separate from the WaPo deal (which, believe me, I have questions about, too). ChatGPT would do this, deal or no deal, because it just completely makes stuff up. Here's the same prompt but with Maggie Haberman, even though NYT has no OpenAI deal and is suing

Super important info for writers to understand. If you have any remotely publicly used name (and sometimes even if you don't), it will do this. That said, specifically citing this garbage as your style, plus the byline, should certainly should open it to defamation.

Also @drewharwell.com re: your contract, do you have any recourse? I know some Atlantic writers were saying like "welp, technically they can do whatever they want with our writing". That's obviously not the case for authors, for example, who explicitly retain copyright. Does your contract allow...

...them to sell your work in this way without compensation or explicit permission? I really hope some writers in cases like this will sue regardless, because clearly even if there is an argument that it's legal the contract was not signed with that understanding.

Even if it didn't make it up, it might use free versions of her articles available elsewhere, quoted excerpts, or scrapers may have intentionally/unintentionally gotten around a paywall. I assume you could ask for "random 2000's era blogger." Might even be able to ask where it pulled examples from.

If I am interpreting the photos correctly then ChatGPT included a fake headline and byline in its response to the style-of-Gillian prompt but not in the response to the style-of-Haberman prompt. I wonder what caused that difference.

Good question. I suspect it's just a weird LLM black-box inconsistency or something NYT / WP domain-specific. It made up a headline and byline for this story, with a similar prompt

The reply follows a statistical path, so if you just hit "Retry" (or "Try again" now) you'd get a different answer Strictly speaking, the model isn't answering the question, it's saying what an answer is likely to look like. But there's some randomization and statistical modelling

It's a statistical model. It doesn't produce the same thing every time. Here's a sample for noted Wapo staff writer Morgan Bird. (I am neither noted nor a WaPo staff writer)

yes, outside of alarm over the separate post deal, i’m missing why this is a personal story about her. ChatGpt has long been able to do this for every writer that has been published

I'd be shocked if OpenAI waited until now to add WaPo to their training set. No way they waited two years to ask for permission; they didn't give anyone else that courtesy lol

(from alt SocMed post) As facetiously seen on Bluesky to comment on ex-WAPO writer's outrage against Shat-GPT. Not to say she's wrong on feeling outrage, but don't think for a moment, with or without legislation, this is NOT going to happen. It's about where we nudge the direction of our "future".

First, what they are trying to do is ethically concerning. Second, AI is no where near ready for what they are trying to do. I try and use AI technology whenever i can to see with my own eyes what is real and what is hype. Sadly the hype prevents us from meaningfully engaging the real

Northeastern is partnering with "Claude" AI school wide. Will students no longer see primary sources, just this blendered info programmed & filtered by the military-govt-tech-AI-surveillance complex?

Orwell and Kafka warned us about giving artificial clowns 🤡too much power …‼️❌🎪❌‼️

Jozef Imrich@jozefimrich.bsky.social

Saving the free press - before it is too late … 🔥 As threats to journalism grow louder across the U.S., a coalition of journalists, nonprofit leaders and press advocates has united to push back urgently. Press Freedom United, launched earlier this year, is more than a campaign.

I think anyone who has ever taken joy in writing understands your horror. I'm so sorry this has happened, but I think this is something that cannot be stopped. It's hard to see a craft you take so much pride in being commoditized. I try to explain this to people around me, but they don't get it.

I'm confused, how is this any different than you asking a human who's familiar with your work to write an article in your voice? WaPo didn't do this, or publish it. You asked for an impersonation and got one. If they actually try to pass off an impersonation as your writing, you'll have a case!

This reminds me of when actor Crispin Glover sued Universal studios for using his likeness in Back to the Future II. They are essentially plagiarizing and impersonating you.

For what it’s worth, and writing as a lay person not a lawyer, in the UK you would probably have a case over your moral rights of integrity and being erroneously identified as the author.