OpenAI came after our domain because we use GPT in it

235 points by decide1000 ↗ HN
Tldr; OpenAI registered the brand GPT and wants us to take down our gptsomething.io domain. The domain was a blank page.

OpenAl brand guideline outreach

Dear Team, We are contacting you on behalf of our client OpenAl, Inc. ("OpenAI"). Our job is to work with developers to ensure that they are marketing their products in accordance with OpenAl's brand guidelines https://openal.com/brandand in a way that doesn't create confusion for the consumer.

Today, we're writing regarding your use of the mark "GPT" in connection with your GPT' product.

https://branddb.wipo.int/en/quicksearch/brand/CH502023000001493? by-brandName&v=GPT&rows=30&sort=score%20desc&start=0& -1682836548861&fcstatus-Registered&fcdesign ation=CH&I=Q

Stating "GPT" is inaccurate and may imply a partnership or endorsement where there isn't one. If your project uses GPT-3, GPT-4, or ChatGPT you may choose to say in the Product description (not in the Product/Site name) that it is "Powered by GPT-3" or "Powered by GPT-4" or "Powered by ChatGPT" and/or "Powered by DALLE" as applicable. We do not permit model names in products/site titles because there is concern that it can confuse end users, and it also triggers our enforcement mechanisms. It's important to identify the GPT-3, GPT-4, or ChatGPT (or DALL-E) model specifically, as opposed to just referencing GPT.

Please reply to this message by 05.10.2023 to let us know that you received it and intend to make the applicable changes. If you have any questions, please contact us directly at openal@brandshield.com.

Sincerely,

BrandShield Ltd.

174 comments

[ 3.0 ms ] story [ 252 ms ] thread
OpenAI is quickly becoming a villain in order to protect their moat and ensure others can't catch up to them.
Hey, it's not their fault, closedai.com [1] was already taken.

[1] Apparently it's really a thing, funny, https://closedai.com

The same guy seems to also have closeai.com, and it eventually links to superangels.co, which actually redirects to a Google search for superangels.com. What a strange little corner of the internet.
Maybe publish some material related to GUID partition tables and how they work. Since I believe they own the trademark for GPT in relation to language models (yeah it sucks) it may be the only way to use your domain.
gptchat.io: The Internet's #1 source for GUID Partition Table talk.
Finally, somewhere that is covering the issues I care about but that the mainstream media refuses to cover.
Ages ago I wrote a parody web page about BSML, and some Bioinformatic Sequence Markup Language shovelers complained to me about it.

You will never guess what excrement of theirs that I called them on.

https://www.donhopkins.com/home/catalog/text/bsml.html

A manufacturer of solid lubricants made from a mixture of polymers, oils, and other additives also complained to me about my web site about Microsoft Monopoly. I wouldn't sell, but they grabbed it up after I let it expire.

https://donhopkins.com/home/Micropoly/

https://www.micropoly.com/

>Since I believe they own the trademark for GPT in relation to language models

EinsteinGPT and BloombergGPT exist, so I'm left wondering if that's true.

Personally excited for ChatMBR.
the rare actual lol from me on this one, well done.
"Open" AI
I'm not really well-versed in this kind of thing to make any particular comments, but this made me rage:

>Please reply to this message by 05.10.2023 to let us know that you received it and intend to make the applicable changes.

How effing arrogant do you have to be to say something like this? No "please respond and we can talk about it", it's "please confirm you know we're the big guys, and you're going to do what we say".

But nice of them to give the site four and a half months to think about it.
That deadline passed 12 days ago
irb(main):008:0> Date.parse("05.10.2023") => #<Date: 2023-10-05 ((2460223j,0s,0n),+0s,2299161j)>

Should be for October 5, according to that format.

Yeah, if you insist on using your weird American date format, use "May 10th, 2023" or at least "05/10/2023" to make it less ambiguous, using dots implies DD.MM.YYYY the same way as using dashes implies YYYY-MM-DD...
> use "May 10th, 2023" or at least "05/10/2023" to make it less ambiguous, using dots implies DD.MM.YYYY the same way as using dashes implies YYYY-MM-DD...

For most of the world, using slashes also implies DD/MM/YYYY. For instance, I would write today's date as 17/05/2023.

The good news is on a dozen days during the year the US and the ROW are on the same page (MM/MM/YYYY or DD/DD/YYYY if you prefer).

Somebody should check if those days are generally associated with fewer accidents, higher stock returns and a pronounced sense of global peace and harmony

Impossible, what madman would use a date-format that goes $MONTH.$DAY.$YEAR that basically have a random order instead of $DAY.$MONTH.$YEAR or $YEAR.$MONTH.$DAY? No one would be that crazy, it's too ambiguous to be useful.

It's clearly intended to be 5th of October, 2023.

(This may or may not be sarcasm, in case people may not or may misunderstand)

Maybe that has something to do with feet and inches instead of metric.
I'm not american, but they always seem to do month.day.year for some bizarre reason
i'm aMerican and I always use year-month-day
Same here but I'm also a programmer, YYYY-MM-DD
Right ISO 8601... in SQL Server it's the unambiguous format
I'm an American who agrees. YY/MM/DD is the only reasonable order.
It's because we write the numerals in the same order as the longer form: May 10th, 2023 becomes 5/10/23.
Not American, and maybe it's a cultural difference but in England it's much more common to say "the 10th of May".

Then again, we write our dates in a more sensible manner. I suppose it's time to link to the old sapir whorf hypothesis[0] to explain the difference.

[0] https://en.m.wikipedia.org/wiki/Linguistic_relativity

> Then again, we write our dates in a more sensible manner.

yawn

Yeah we write our dates in a more sensible manner too. And also we drive on the correct side of the road and imperial units of measurement are superior.

I have been in the US for 30 years but metric is way better... all that nanometer stuff in your computer is metric not imperial as well

for example a 10cm cube filled with water weighs 1 kilo and is 1 liter, boils at 100 degrees. In the US you deal with cups pints, quarts, gallons... and you can't even convert easily between dry and liquid stuff

I do agree that steering wheels should be on the left :-)

> all that nanometer stuff in your computer is metric not imperial as well

Electronics is full of imperial standards.

I was being sarcastic because none is better, it's what we're used to. I can do measurements in both imperial and metric, but because imperial is widely used in MY country, I find it "superior", same with everything else.
I'm an American but I'd say metric is superior, even though I'm not used to it. It has more consistency and is easier to actually do science with.
Decimals proceed in order of refinement to smaller increments. Years are long, months are collections of days, and days are collections of hours, and so forth.

By that logic, all formatted dates should follow a progression of precision. This is, in fact, a standard: https://en.wikipedia.org/wiki/ISO_8601

Also agreed, and what i personally use.
I thought the Americans use MM/dd/YY. As soon as I see a dot "." instead of "/" I assume a "normal" day-month order.
I wouldn't expect them to be that precise, especially when they misspell their client's name as OPENAL
I'm not american, but they always seem to do month.day.year for some bizarre reason

It's one of those things that America inherited from England, and then when England changed its method, the Americans saw no need to change because it wasn't part of England anymore.†

If you go into antiques stores in England, you will find that with items that are engraved, the older it is, the more likely it will have the now-American date format.

There was an item on Antiques Roadshow a couple of months ago that was a gift from Queen Victoria to someone that had an engraving that included the American date format.

† This is true for a surprising number of minor differences between America and England. In life, it's often better to understand the history of things, rather than just ignorantly write off a group of people as bizarre.

I think that's because that's how the full date is spoken. It's "September Eleventh" not "Eleventh of September". If someone said "My birthday is twenty-seventh of November" I would wonder if they're a synth.
But outside US, '11th of September' IS how it's said
I'm American (South American, to be more precise) and we do DD/MM/YYYY around here.
And that's why I always write month as a word
I wonder how that would play out in court. Is it a valid defence to say that you responded a few months late because your local jurisdiction uses DD-MM-YYYY?
The letter is a demand for action that assumes a conflict exists and the client is correct. It is standard. The recipient can acknowledge and agree or acknowledge and disagree.
AFAIK this is fairly standard lawyer-speak.
When did you register it first?
Too bad this domain is already taken gpt-shirts.com.

"They trademarked GPT and all I got was this T-shirt"

Or GPT-shirt, as in generic plain t-shirt.
> openal.com

Open Audio Library?

Good catch, I did not notice that "I" and "L" are intermixed in the letter. In particular, their link uses "L" instead of "I".
So it's not even from OpenAI, it's a fishing scam
I always make sure to write Al when talking about AI. As in Weird Al Yankovich.
I used OCR on a screenshot of the email. That's probably the reason of the mixup.
I've had the fortune of receiving an even more obnoxious threat from another big co., fortunately my lawyer helped me make them go away by advising me to stand my ground and just tell them repeatedly how their claims are completely unfounded.

Your case might be different, but you may want to ask for professional advice as well.

Everybody thought “Windows” was too generic, but Microsoft successfully defended the trademark. It seems very similar to OpenAI’s claim here.

The formulation that Microsoft allowed was “N for Windows”, but not “WindowsN” or similar. (For example, SuperFixer 1.0 for Windows instead of WindowsSuperFixer 1.0)

Drake also supposedly trademarked "God's Plan" (for use in clothing).

I wonder how God feels about that one.

"for all have sinned and fall short of the glory of God" (Romans 3:23 NIV) ;)
https://en.wikipedia.org/wiki/Generative_pre-trained_transfo...

I think this should not be allowed for OpenAI to own this term

Isn't your link more of a reason pro OpenAI's trademark claim?

>The concept and first such model were introduced in 2018 by the American artificial intelligence organization OpenAI

https://webcache.googleusercontent.com/search?q=cache:_aUoGH...

https://www.bizjournals.com/sanfrancisco/inno/stories/news/2...

**From the article:

"...the actual phrase GPT, standing for generative pre-trained transformer, is a broad term describing the AI technology. The term was first introduced in 2018 in a research paper published by OpenAI, but has since been used to describe any AI model that meet certain criteria.

This broad use could cause problems for OpenAI as the U.S. Patent and Trademark Office says it can refuse to trademark a term that is 'merely descriptive.'"

Gentle reminder laws are fake and act only as a differentiator between having or not having money/power: 'Fashion designer Marc Jacobs and Ohio State University separately filed applications with the U.S. Patent and Trademark Office in 2019 to trademark [the word] “the”' [1].

[1] https://www.fastcompany.com/90763993/why-marc-jacobs-and-ohi...

Of all the "IP" laws, trademark laws actually serve a necessary role in society, and their applicability is actually pretty well limited. Trademarks exist to make it clear who is actually selling a piece of merchandise you are buying.

Trademarking a common word is not usually a problem, as the trademark only applies within a specific industry. For a famous example, a small IT startup (Apple Computer Company) was allowed to use the same name as one of the biggest and most famous record labels in the world - the Beatles' own Apple Records. Both continue to exist to this day, with their respective trademarks, though of course their relative importance has slightly shifted.

I wonder if my hypothetical sites “The Laughing Pterosaur” at thelaughingpterosaur.com or “OpenPGP Tips and Tricks” at openpgptipsandtricks.com get flagged as well.
IANAL, but make sure you include real pictures of laughing pterosaurs and not generated ones.
It's bad for OpenAI to go after a blank website, which could develop into anything, not necessarily related to their GPT product. Other than that, this is exactly how trademarks are supposed to work. OP should just calmly tell OpenAI that they are not infricting on the trademark, because it's not in the same business domain.

I get that people have a negative attitude towards the company because the "open" in its name basically turned out to be massive bait. But this is business as usual and trademarks exist for good reasons.

(comment deleted)
The problem is precisely the spamming of takedown requests without doing any due dilligence that the target actually does infringe at at all, essentially offloading your enforcement costs to random individuals. IMO reapeated offenses of this kind of behavior should result in complete loss of the trademark.
The web site being blank works in OpenAI favor as cybersquatting works against the domain owner.
OP should just calmly tell OpenAI that they are not infricting on the trademark, because it's not in the same business domain.

OP should not communicate with OpenAI without a lawyer at this point.

The one exception is if this vaguely threatening letter turns into a formal cease and desist, and OP cannot budget for a lawyer.

OP, did you use OCR or find/replace on this email? Every occurrence of "OpenAI" in their message ends in lowercase L, as in OPENAL (with an amusing exception of the name in quotes). That includes the email to reply to.

Also, I'm pretty sure you can't claim trademark infringement for an empty page and a domain made of generic words, because trademarks are limited to context.

(comment deleted)
Yes I used OCR on the screenshot of the email they sent.
That seems like a lot of extra unnecessary effort for copy/paste-able plaintext.
(comment deleted)
Plain Text > Screenshot -> OCR -> Plain Text can be used for avoiding some sorts of identifiers hidden in text.
Yes. It just makes the text harder to read for no good reason. It’s plain-text – not an uploaded Word document or PDF with lots of hidden metadata.
Even plain text can contain hidden metadata.
I think you're confusing text documents like PDF or word .doc(x) type files for "plain text". Plain text doesn't contain hidden anything at all. It's just plain text and nothing else. That's why a plain text editor is completely different from Microsoft Word, or OpenOffice. They both edit textual document types, but a plain text editor doesn't do metadata, fancy fonts, colors, styles, or any of the fun junk you find in PDF and other textual document types. It does only text.
I know what "plain text" is and even so, it can contain identifiable markers, if you want to track down a leak let's say.

"Something 𝗌omething will launched on Novemᖯer 3rd"

This "plain text" string has two markers that aren't actually ascii, `𝗌` and `ᖯ`. By changing the markers for each person you send it, you might be able to track down who is leaking things, if they're careless and just copy-paste stuff that look like plaintext but actually isn't.

Anything that is a homoglyph (https://en.wikipedia.org/wiki/Homoglyph) can be used in this manner

"Plain" text can contain hidden characters such as zero width spaces which won't be visible to someone reading it but could be used for identification purposes.
Funny how "OpenAI" speedran themselves in record time to be the litigious and moat-building villains of AI
You haven’t said what the GPT in your domain name means. Is it actually a reference to AI or is it completely unrelated?
I think Elon said something interesting about OpenAI in an interview he gave yesterday which I didn't perviously understand – as part of their deal with Microsoft Microsoft have right to all of OpenAI's model weights. This means at any time Microsoft has the right to take their models and cut off OpenAI's funding.

I would seem OpenAI now have very little autonomy as a company. Since any IP they currently have can be taken by Microsoft there's very little incentive for other companies to invest. And since OpenAI isn't profitable they are now 100% dependent on Microsoft to continue operating.

I think we may as well consider OpenAI a Microsoft acquisition since this appears to more accurately describe the relationship here.

Perhaps Sam Altman was just too stupid to understand this, but I suspect that isn't the case. It seems he probably knew OpenAI would effectively become a Microsoft research project as part of their deal, and therefore we should probably assume that they will operate with for-profit incentives – and with those of Microsoft's more specifically.

I guess what I'm saying is that this doesn't surprise me, and I'd expect more of this to come.

> as part of their deal with Microsoft Microsoft have right to all of OpenAI's model weights. This means at any time Microsoft has the right to take their models and cut off OpenAI's funding.

How does he know this? Is it because the terms of the OpenAI / MSFT deal are public knowledge? Or is he just making it up?

He was an early investor. $50m. Maybe that gives him special access to docs.
I don’t think Musk is a very reliable narrator in general, and given the history he’s probably particularly unreliable in the context of OpenAI.

As far as I know, the terms of their deal with Microsoft aren’t public, but all the reporting I’ve seen has suggested the opposite - that OpenAI basically kept full control over anything resembling the crown jewels. And why wouldn’t they? It was a frothier market when the deal was made, and it’s not like they couldn’t have raised capital from dozens of other sources.

Do you think Microsoft would actually take a risk like that? They've invested heavily in Open AI, so I'd be pretty surprised if they don't have access to the IP.

Like you said the legal details aren't public, but I can't imagine they don't have access to some of the IP.

Isn't Microsoft sitting on a huge pile of cash? It seems that having a piece of the AI pie if it's truly going to explode as much as everyone seems to be expecting, plus free access to their APIs, would be reason enough.
This would assume they have no strategy and are just hoping Open AI won’t back out of any deal or somehow degrade their product and then they’d be left in the dark.

I would find this hard to believe.

And to add to this, what would be the point of keeping the terms of the deal secret if it's just a simple investment on the part of MS? You'd think "Open" AI would want would to be open about these things.

And as an initial investor in OpenAI I'd assume that Musk would be connected enough to get details about the deal even if those details haven't been made public.

Was that the same interview in which Elon lied through his teeth and claimed that the Allen, Texas shooter did not hold white supremacist beliefs (despite tons of open source evidence and even the confirmation of local police) and that Bellingcat does psyops?
I can't find any reputable sources that confirm he was a white supremacist. Can you post one of the links you have?
>I can't find any reputable sources that confirm he was a white supremacist.

Yep, there's the weasel word, "reputable."

There are plenty of sources, you can simply google it. What sources would you consider "reputable?"

I believe he is said to have nazi tattoos.
I don’t understand why that means he’s necessarily a white supremacist. For one, chances are he wasn’t all that smart. For two - it could just be that he was an anti-Semite, and so he latched on to Nazi iconography.
Why are you referring to Mauricio Garcia, the Hispanic shooter, as a white supremacist? I read through ten articles about him just now - all of them said he was a white supremacist, none of them bothered explaining how a Hispanic man was a white supremacist. Surely in a sane world one of them would at least pay lip service to explaining that to me, but nope.
Have you ever applied for a job in the US, or answered a census here? White is broken down into Hispanic White and non-Hispanic White in our racial categories. That’s about as official as you can get.

Also consider what “Hispanic” means in the first place. Plus, ultimately, anyone can have any beliefs if they really want to. If you’re non-white and try to be a white supremacist, it’s not like the universe will prevent you from doing so.

It would make you a hispanic supremacist or white hispanic supremacist if using the official categories. Using incorrect labelling is done on purpose to keep a false narrative going and attract more attention.
The media not describing a guy (who checks “Race: White, Ethnicity: Hispanic” on the census), who had a swastika and SS tattoo, as a “Hispanic supremacist” in headlines is maybe the weakest media criticism I’ve ever read.

I have the craziest hunch that maybe you aren’t being genuine, but I’ll leave that alone.

100 genuine. I would have expected them to use the common latin or spanish. Not white supremacist. That feeds into a stereotype that doesn't exist here. It's like labeling the Oklahoma city bomber a muslim terrorist after 9/11 to fit in with the stereotyping of the day.
Suuuuure you're "100% genuine" in suggesting that calling a neo-Nazi with a fondness for sharing violent white supremacist rhetoric a white supremacist is like calling Timothy McVeigh a Muslim, and insisting they ought to focus on the neo-Nazi's Spanish name instead of his "right wing death squad" patches. Can't be having the media accept people as white if they have Spanish names, can we?
So isn't the reason that Mauricio Garcia is labelled as a white supremacist because he himself actually wrote and self-identified as one?

If the Oklahoma city bombers did not leave behind any jihadist or pro-Islam writing of any sort and only left behind their white supremacy stuff I don't think they would have been labelled as Muslim terrorists if they did their attack soon after 9/11. The media very likely would have done something more like "white supremacy as displayed by the Oklahoma City Bombing is similar to the jihadist doctrine displayed by Al-Qaeda". Which is to say the media would have made links between the two (some spurious, some genuine) without actually saying the Oklahoma City Bombing was done by Muslim terrorists. This is so obvious to me to the point that I am beginning to feel empathetic to commenters doubting your being genuine - as much as I hate to admit it since I think trusting others are arguing in good faith is generally a better norm.

Whether the 9/11 era or now when someone commits a crime and actually leaves behind writing where they admit their beliefs, and motivation this is what they get labelled as. Why are you surprised when someone like Mauricio Garcia who admits white supremacist beliefs is then labelled as one by the media? Or is your claim that actually his writing is not white supremacist?

Are white supremacists generally believed to be referring to Hispanics as supreme? If not, I contend it is reasonable for me to have assumed at least one of the ten reports I read would spend a moment addressing this dissonance.
Models are worthless without reproducible training anyway. What good is a point in time model if you can't update it going forward.
It's almost true that they have no moat, or have a small one. Acquisition is the best option for them and the time window for that is closing as it won't be long until every company can afford to create their own gtp4. The value of those weights is going down fast, there are now hundreds of LLMs of medium size, all it takes is for any company to change the parameter size and retrain.
My understanding is that Microsoft gets exclusive licensing to (just) the "GPT" family of technologies. This is why Sam Altman has been hoping to get away from transformers and attention-based methods.

In any case, I'm confident that their researchers could come up with something else if Microsoft truly decided to "take them over". Furthermore, it seems like it's still a mutually beneficial arrangement.

Sorry to be pedantic, but I'm failing to see how this is related to the OP?
Their statement is plain ridiculous scare tactics. A trademark registration is not a monopoly on any and all use of that combination of letters.

If you are in Switzerland, you might have a right to use the domain because you registered it before the trademark was registered. Ask a trademark attorney about application of Art. 14 of the Swiss Trademark Code.

Plus, they may have no right to enforce anything at all if you do not use the domain for their registered goods/services. Again, ask a trademark attorney.

Disclaimer: I'm not an attorney and this is not legal advice.

1. Can you claim that GPT are your initials? (Is your name George Patrick Tanner? Gina Parker Taylor?)

2. If you are using it for AI, then you're not going to be able to show that "GPT" is a term of art predating their company. The first appearance of the term for AI is from a paper by OpenAI.

3. Check the dates. Did you register your domain name before they filed their patent/trademark? (They filed in on Feb 3, 2023.) If so, then you have an argument for prior art.

4. Don't make an offer to sell (even in jest) without talking to an attorney. They could interpret it as a bad-faith negotiation and take the domain.

5. Do you plan to use the web site for something unrelated to AI? Information about GUID Partition Tables? Goniopora Toxin? General-purpose technologies? Generalized probabilistic theory? Grounded practical theory? Anything like this doesn't compete with their trademark.

> They filed in on Feb 3, 2023.

IANAL but that's just the swiss trademark. The US trademark is from Dec 22, 2022. Serial number 97733259.

It would'be been so apt had it been Dec 22, 2012.
>> 1. Can you claim that GPT are your initials? (Is your name George Patrick Tanner? Gina Parker Taylor?)

That didnt work so well in the past:

https://en.wikipedia.org/wiki/Microsoft_v._MikeRoweSoft

> A settlement was eventually reached

Or in other words, we didn't get to find out how well it would work.

> Robertson opined that – had legal proceedings ensued – Rowe would have made a strong argument for keeping his domain, as he was using his real name and was not claiming to be affiliated with Microsoft.

>> Or in other words, we didn't get to find out how well it would work.

What small company or individual has the tens of thousands of spare dollars to actually litigate/defend? It doesn't matter if a settlement was reached or whether a strong argument was made --- if you have to spend 10k, or 15k, or 20k to go retain lawyers...it hasn't worked out well!

I'd let em know that you have plans for the domain but they could have it for $SILLY_CASH_AMOUNT. Maybe get chatGPT to polish up the language.
It's a scam to get a hold of your domain name. They can't own the term GPT.
GPT meant, "Get Paid To" in the incentive advertising world. Maybe your product was using this acronym?

These robo-takedown outfits are a form of spam in my view. If you are publishing large sites or include user generated content you've probably run into them more than once. From what I've seen claimed as infringing it is obvious that there is no human review process. Compare to the expense of filing court actions for bulk frivolous claims.

Typically they can be ignored if your hosting provider is game. Their first step will be to try to harvest your data by provoking a response. If you do feel compelled to respond, be sure to run it by someone with legal expertise. Anything you say can be misconstrued in the worst possible way.

At the end of the day, these companies want to produce metrics to say, "We defended your brand from N number of verified infringement incidents this quarter. Based on the escalating threat volume, we recommend giving us more money going forward..."

IANAL, but I would avoid interacting with these 3rd party trolls. If they truly want to send a demand letter, ChatGPT's lawyers should send it by courier and obtain receipt of your signature. Until then I wouldn't take it seriously. This doesn't even rise to the level of a genuine legal threat or registered demand letter. It is more of a scare tactic.

IANAL, and if you want to stand up to them you probably need one.

However, note that they do not use the word trademark anywhere. That's because they do not yet own the trademark on GPT [0], it's still awaiting examination. I'm not an expert here, but this letter feels to me like they're starting to try to intimidate people into following their guidelines as prep work to ensure that it doesn't become generic before they even get the trademark.

Again, IANAL and I'm not giving advice, but this is not a trademark infringement warning because there is (as of yet) no trademark.

[0] https://tsdr.uspto.gov/#caseNumber=97733259&caseSearchType=U...

This is my personal opinion and not legal advice. IANA trademark practitioner.

Before the EUTM [0], the EU trade mark office, they're currently in opposition proceedings with no oppositions registered.

AFAIK the term GPT is entirely descriptive (so not distinctive) and so is unsuitable for use as a trademark. It should be refused, but I wouldn't expect trademark examiners to have heard of generative-pretraining, nor that this can be applied to transformers to make a gpt. Unless someone tells them then I expect the application by OpenAI would be allowed.

That said, OpenAI can use the term "trademark" if they're using that term as a mark of origin for goods or services, it doesn't need to be registered first (at which point it becomes an RTM and can use the (R) symbol). Indeed you can register a mark as having acquired distinctiveness through your prior use as a trademark ("acquired distinctiveness").

[0] https://euipo.europa.eu/eSearch/#details/trademarks/01883665...

> Unless someone tells them then I expect the application by OpenAI would be allowed.

How does one tell "them" and why don't you?

How do I file an opposition? Oppositions against EU trade mark applications have to be filed in writing within a period of 3 months following the publication of the EU trade mark application in the EU Trade Marks Bulletin. For your convenience, oppositions can also be filed using the online form.

https://euipo.europa.eu/ohimportal/en/opposition#:~:text=How....

Someone should really use ChatGPT to write an opposition letter and submit for the lols

I think "ChatGPT" is distinctive to their specific product, but not just "GPT" in general.
Actually, what I think I was looking for was "observations", oppositions is for other parties with standing (eg other traders, other trade mark owners) I think. Whilst observations, also part of the adversarial processing, https://guidelines.euipo.europa.eu/1803468/1786255/trade-mar..., are for third parties.

Opposition has a fee, and it's probably not appropriate for me personally to be involved (I work in IP, not in TMs). Maybe I'll look at anonymous submissions under Art.5 EUTMR.

(That link says it is obsolete, I'm assuming that the procedure has just been updated.)

IANAL but also there is a specific list of classes (categories) that have to be selected from when seeking to register a trademark, design mark etc and afaik protection then applies only within those categories.

In this case they have applied for Nice classes 9 and 42.

I'm not sure, but also seem to recall that protection within those classes would be further limited to the scope of what they covered in the description they provided for each class during application.

Anyways, point it is that it's in any event really hard to see how having a fairly nondescript domain name with a blank web page could ever be construed as purveying anything in either Nice class 9 or 42.

(Eg, for all we know, such domain name might be referring to some new kind of industrial oil or lubricant, which would be Nice class 4...)

It's also useful to consider that there are contexts where the law can negate the validity of a trademark based on the registrant's lack of actively "practicing" it (including during the period between application and it being granted), or protecting it. So there's an inherent incentive for companies to be a bit "trigger happy" in staking their claim to "practicing" the band as well as acting to protect it. Although it does seem very preemptive in this case.

You don't need to be a lawyer to know that you have to defend a trademark else you risk losing it.

A similar law exists for property.