Daphne Keller
@daphnek.bsky.social
Director of Platform Regulation, Stanford Law School LST Program. Former Google (2004-2015) Legal Director for Web Search, Speech and Intermediary Liability Issues.
May be cranky.
https://law.stanford.edu/daphne-keller/
May be cranky.
https://law.stanford.edu/daphne-keller/
Here @pjleerssen.bsky.social and I are pointing this out in 2020.
It is important to keep saying it. Maybe someone could say it to law review editors and the publishers of tell-all books. I'm sure they'd be very receptive.
papers.ssrn.com/sol3/papers....
It is important to keep saying it. Maybe someone could say it to law review editors and the publishers of tell-all books. I'm sure they'd be very receptive.
papers.ssrn.com/sol3/papers....
November 23, 2025 at 7:06 PM
Here @pjleerssen.bsky.social and I are pointing this out in 2020.
It is important to keep saying it. Maybe someone could say it to law review editors and the publishers of tell-all books. I'm sure they'd be very receptive.
papers.ssrn.com/sol3/papers....
It is important to keep saying it. Maybe someone could say it to law review editors and the publishers of tell-all books. I'm sure they'd be very receptive.
papers.ssrn.com/sol3/papers....
I've given up on pointing this out, but it is 100% true. The list of "new discoveries" about platforms that I literally filed public affidavits about in 2010 or testified on live TV about in 2012 is... really long.
We could skip a bunch of bullshit by looking at what is already public.
We could skip a bunch of bullshit by looking at what is already public.
The biggest source for any DSA investigation into addictive design features might not be platforms' risk assessments, but the more than 6 million documents that are part of the pretrial discovery process in the US www.bloomberg.com/news/article...
‘Massive Legal Siege’ Against Social Media Companies Looms
Thousands of plaintiffs’ complaints, millions of pages of internal documents and transcripts of countless hours of depositions are about to land in US courtrooms, threatening the future of the biggest...
www.bloomberg.com
November 23, 2025 at 7:01 PM
I've given up on pointing this out, but it is 100% true. The list of "new discoveries" about platforms that I literally filed public affidavits about in 2010 or testified on live TV about in 2012 is... really long.
We could skip a bunch of bullshit by looking at what is already public.
We could skip a bunch of bullshit by looking at what is already public.
I get into some of that here, more in the pending post. For scraping, the Commission has now said 3x that a TOS prohibition = failure to "give access". Which is a kind of funny formulation, but the reality is TOSes deter research, so it's important.
verfassungsblog.de/dsa-platform...
verfassungsblog.de/dsa-platform...
Using the DSA to Study Platforms
verfassungsblog.de
November 23, 2025 at 6:56 PM
I get into some of that here, more in the pending post. For scraping, the Commission has now said 3x that a TOS prohibition = failure to "give access". Which is a kind of funny formulation, but the reality is TOSes deter research, so it's important.
verfassungsblog.de/dsa-platform...
verfassungsblog.de/dsa-platform...
I do think that some drafters of 40.12 were imagining CrowdTangle (hence R98), and some imagined API use. For both, it makes sense for platforms to "give access" despite data being "publicly accessible" already. So the drafting is nutty, but you can strain to find a sensible meaning...
November 23, 2025 at 6:54 PM
I do think that some drafters of 40.12 were imagining CrowdTangle (hence R98), and some imagined API use. For both, it makes sense for platforms to "give access" despite data being "publicly accessible" already. So the drafting is nutty, but you can strain to find a sensible meaning...
Big parts of Art 40, including 40.12, were added by the Council draft and resolved in trilogue. So the whole time there was a public debate, or Parliament weighing in, etc. this language simply wasn't on the table. It's no surprise it's somewhat incoherent.
November 23, 2025 at 6:52 PM
Big parts of Art 40, including 40.12, were added by the Council draft and resolved in trilogue. So the whole time there was a public debate, or Parliament weighing in, etc. this language simply wasn't on the table. It's no surprise it's somewhat incoherent.
I agree with this as the bottom line for scraping. Though then there are serious limits from 40.8 on how that data can be used, it's not a free-for-all.
November 23, 2025 at 6:51 PM
I agree with this as the bottom line for scraping. Though then there are serious limits from 40.8 on how that data can be used, it's not a free-for-all.
"incredible smell" is a great way to put it :)
November 23, 2025 at 6:49 PM
"incredible smell" is a great way to put it :)
It just leads to weird, arbitrary answers because no one was planning for this stuff when drafting.
November 18, 2025 at 7:37 PM
It just leads to weird, arbitrary answers because no one was planning for this stuff when drafting.
I think it makes sense for private FB posts to be DSA regulated, so they must be "publicly disseminated."
But it doesn't make sense for researchers to see them, so they are not "publicly accessible".
That's super outcome-oriented. But parsing the words carefully doesn't lead to better answers.
But it doesn't make sense for researchers to see them, so they are not "publicly accessible".
That's super outcome-oriented. But parsing the words carefully doesn't lead to better answers.
November 18, 2025 at 7:37 PM
I think it makes sense for private FB posts to be DSA regulated, so they must be "publicly disseminated."
But it doesn't make sense for researchers to see them, so they are not "publicly accessible".
That's super outcome-oriented. But parsing the words carefully doesn't lead to better answers.
But it doesn't make sense for researchers to see them, so they are not "publicly accessible".
That's super outcome-oriented. But parsing the words carefully doesn't lead to better answers.
I think there is no one right dividing line. But for any given legal obligation, different dividing lines may make sense.
November 18, 2025 at 7:36 PM
I think there is no one right dividing line. But for any given legal obligation, different dividing lines may make sense.
Oh, good point! You are so good at spotting general monitoring issues.
November 18, 2025 at 7:32 PM
Oh, good point! You are so good at spotting general monitoring issues.
The same goes for "access", though I think that word only became quite so freighted with different disputed legal meanings in the digital era. But "public" has been a sort of linguistic placeholder for complicated values forever.
November 18, 2025 at 7:32 PM
The same goes for "access", though I think that word only became quite so freighted with different disputed legal meanings in the digital era. But "public" has been a sort of linguistic placeholder for complicated values forever.
One of my big points in this scraping project is that "public" means different things in different laws (news reporting, insider trading, copyright, etc.). There is no point trying to force those to all mean the same things, because the values and purposes of the laws lead to different meanings.
November 18, 2025 at 7:30 PM
One of my big points in this scraping project is that "public" means different things in different laws (news reporting, insider trading, copyright, etc.). There is no point trying to force those to all mean the same things, because the values and purposes of the laws lead to different meanings.
Yeah. But it's probably unavoidable. The diversity of kinds of communication and information online was never going to break neatly into categories, even with as many words and categories as the DSA came up with.
November 18, 2025 at 7:04 PM
Yeah. But it's probably unavoidable. The diversity of kinds of communication and information online was never going to break neatly into categories, even with as many words and categories as the DSA came up with.
Or, as I said somewhere else in this thread: I'm on team teleology. But the words are making that hard.
November 18, 2025 at 7:03 PM
Or, as I said somewhere else in this thread: I'm on team teleology. But the words are making that hard.
Hmm I think neither "pub disseminated" nor "pub accessible" data includes any internal metrics.
Internal data is what 40.4 is for. External data is what 40.12 and "pa" is for. Classifying services for regulatory obligations is what "pd" is for. I feel like that matters more than the words, alas.
Internal data is what 40.4 is for. External data is what 40.12 and "pa" is for. Classifying services for regulatory obligations is what "pd" is for. I feel like that matters more than the words, alas.
November 18, 2025 at 7:02 PM
Hmm I think neither "pub disseminated" nor "pub accessible" data includes any internal metrics.
Internal data is what 40.4 is for. External data is what 40.12 and "pa" is for. Classifying services for regulatory obligations is what "pd" is for. I feel like that matters more than the words, alas.
Internal data is what 40.4 is for. External data is what 40.12 and "pa" is for. Classifying services for regulatory obligations is what "pd" is for. I feel like that matters more than the words, alas.
So I'm not sure there is a right answer. But I very much think the DSA drafters expected FB users to have rights to appeal, got to ODS bodies, etc. and for FB to accept trusted flagger flags for *all* posts, including private ones, on the main FB service. Meta didn't even bother arguing otherwise.
November 18, 2025 at 6:55 PM
So I'm not sure there is a right answer. But I very much think the DSA drafters expected FB users to have rights to appeal, got to ODS bodies, etc. and for FB to accept trusted flagger flags for *all* posts, including private ones, on the main FB service. Meta didn't even bother arguing otherwise.
I think @gateklons.bsky.social and I may see this a little differently. I think there is a gradient, as someone said, and lots of things are in practice arguably public or non-public.
The DSA does a terrible job defining any of this, and scrutinizing its language often just leads to snarls.
The DSA does a terrible job defining any of this, and scrutinizing its language often just leads to snarls.
November 18, 2025 at 6:53 PM
I think @gateklons.bsky.social and I may see this a little differently. I think there is a gradient, as someone said, and lots of things are in practice arguably public or non-public.
The DSA does a terrible job defining any of this, and scrutinizing its language often just leads to snarls.
The DSA does a terrible job defining any of this, and scrutinizing its language often just leads to snarls.
Or in a single page of text and images, some of it may be legal because of fair use and other parts because of a license and still other parts because of de minimis use. You have to take things one little piece at a time to apply the law.
November 18, 2025 at 6:49 PM
Or in a single page of text and images, some of it may be legal because of fair use and other parts because of a license and still other parts because of de minimis use. You have to take things one little piece at a time to apply the law.
I find divvying up different functionalities for different legal obligations very intuitive, which is probably because I think in terms of litigation. Like, in a copyright case, only the part of the service that makes a copy (or other relevant act) matters.
November 18, 2025 at 6:48 PM
I find divvying up different functionalities for different legal obligations very intuitive, which is probably because I think in terms of litigation. Like, in a copyright case, only the part of the service that makes a copy (or other relevant act) matters.
Groups, messages, and individual posts distributed to limited contacts are definitely three different categories. Each deserves its own analysis.
But I don't think the platform's choice to separate things into different apps should be dispositive.
But I don't think the platform's choice to separate things into different apps should be dispositive.
November 18, 2025 at 6:43 PM
Groups, messages, and individual posts distributed to limited contacts are definitely three different categories. Each deserves its own analysis.
But I don't think the platform's choice to separate things into different apps should be dispositive.
But I don't think the platform's choice to separate things into different apps should be dispositive.
Here is what the Commission said about FB messages in its VLOP designation.
November 18, 2025 at 5:12 PM
Here is what the Commission said about FB messages in its VLOP designation.
I am super interested in this and every other current scraping issue (but cannot possibly keep up with it all!).
Whatever the legal defense for this scraping, it can't be Article 40.12 since Discord isn't a VLOP. Scraping has been in weird legal limbo for years and AI is making that limbo change.
Whatever the legal defense for this scraping, it can't be Article 40.12 since Discord isn't a VLOP. Scraping has been in weird legal limbo for years and AI is making that limbo change.
November 18, 2025 at 5:11 PM
I am super interested in this and every other current scraping issue (but cannot possibly keep up with it all!).
Whatever the legal defense for this scraping, it can't be Article 40.12 since Discord isn't a VLOP. Scraping has been in weird legal limbo for years and AI is making that limbo change.
Whatever the legal defense for this scraping, it can't be Article 40.12 since Discord isn't a VLOP. Scraping has been in weird legal limbo for years and AI is making that limbo change.
VLOPs are not allowed to prohibit eligible 40.12 researchers from scraping. If they could, Art. 40.12 would be totally toothless. The Commission has said TOS prohibitions on scraping violate 40.12 in three investigations now.
November 18, 2025 at 4:58 PM
VLOPs are not allowed to prohibit eligible 40.12 researchers from scraping. If they could, Art. 40.12 would be totally toothless. The Commission has said TOS prohibitions on scraping violate 40.12 in three investigations now.