Daphne Keller

@daphnek.bsky.social

13035 Followers 173 Following

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/

Daphne kisses Charlotte on the cheek, Charlotte finds the situation questionable
“4 Requests, 3 Rides… A Law Enforcement Operations Simulation”
X post from age assurance vendor Tony Allen describing meetings in Jakarta. Seoul, Tokyo, Kuala Lumpur, and Dubai
X post from age assurance vendor Tony Allen describing meeting at UK’s Jakarta embassy
500 server error Nous rencontrons un problem technique. You can hear your demand.
Color coded concentric circles showing data transmitted from Insta to a user's device, versus what is available through two research tools
It did not properly assess how the design of its service - including recommender systems and product promotion programmes by affiliated influencers - could amplify dissemination risks of illegal products.
It seriously underestimated how often EU consumers are likely to encounter illegal items. Evidence from a mystery shopping exercise included in the Commission's investigation shows that a very high percentage of the selected chargers failed basic safety tests, while a high percentage of tested baby toys posed safety risks of medium to high severity, as they contain chemicals exceeding legal safety limits or pose suffocation hazards due to detachable parts
Administration officials have in recent days considered creating a roughly $1.7 billion fund to compensate political allies, but not Mr. Trump directly, who say they were wronged by the Biden administration. That fund appears to be part of private deal, reached outside the purview of the court, to resolve both Mr. Trump’s I.R.S. lawsuit and his separate administrative claims against the Justice Department, according to people familiar with it who described it to The New York Times last week. The move by Mr. Trump was a remarkable end-run around the legal system, effectively stripping Judge Kathleen M. Williams, who has been overseeing the case in the Southern District of Florida, of her normal role in approving a formal settlement agreement. By dismissing the case in its entirety, Mr. Trump essentially freed his hand to reach a deal with administration officials without any judicial oversight.
NYT Homepage link to “The Shared Feeling of Being Harvested by the Future” by Yi-Ling Liu, with anxiety-inducing illustration of uniformed delivery people nervously glancing over their shoulders in Tron-like virtual space full of surveillance devices.
404 error from poetryfoundation.org, reading: Oh no! Page not found (404 error) Error is boundless. Nor hope nor doubt, Though both be groundless, Will average out. BY J.V. CUNNINGHAM, FROM "MEDITATION ON STATISTICAL METHOD" We're sorry, but we can’t find the page you’re looking for. It may have been moved or the address misspelled. But don’t give up! Would you like to go to the homepage or browse poetry ? Or try a keyword search.
Proofreaders mark for inserting an emdash
The weeks-long investigation, details of which have not previously been reported, uncovered no records indicating censorship, according to two people familiar with the matter, who spoke on the condition of anonymity for fear of retribution. “There is no evidence that Member States of the European Union are overreaching the DSA to censor and criminalize online content,” they wrote in conclusion.
Decisions were being made to "maintain a strong relationship" with political figures to avoid threats of regulation or bans, not because of the risks to users, the TikTok staffer said.
Yahoo! France Real Lessons Countdown Lesson #3: If you can geotarget ads, that makes it hard to tell national courts that you can’t geotarget legal compliance.
To make legal questions manageable, let’s simplify by generally assuming A single AI company does every step (collecting, training, making available for users to create outputs, hosting outputs) Training data is scraped without permission from publicly accessible webpages The website operator had the legal right to publish the content in the first place Outputs resemble inputs to varying degrees Prompts for outputs come from third party users (not the AI company) “Draw me a bug-eyed orange cartoon cat that likes lasagna” “Make me an audio recording of Gavin Newsom accepting a bribe” IRL many of these things will not be true, and legal analysis will get harder as a result.
X claims 286k moderation acts against ILLEGAL content
Zuckerberg said increasing engagement on Instagram was not a company goal. The judge threatened to hold anyone recording with AI glasses in contempt.
The FTC unconstitutionally targeted NewsGuard for its First Amendment activity, including by conditioning a merger last fall between advertising companies Omnicom and IPG on a prohibition against the new conglomerate using news rating services when determining where to buy ads — drafted and amended to ensure the ban would prohibit access to NewsGuard’s ratings.
This screenshot has text (copied below) and an image of more text, which I can't copy. D. The DSA is the culmination of the European Commission’s campaign to achieve global online narrative control. The Hate Speech and Disinformation Codes were the precursor to the formally binding digital censorship requirements under the DSA. European Commission officials have repeatedly confirmed as much. Prabhat Agarwal, the European Commission’s top career official responsible for implementing the DSA, said during a March 2024 event that the Hate Speech and Disinformation Codes were intended to inflict “reputational damage” on platforms that were not censoring enough content to placate the European Commission.133 The DSA, he said, gave regulators something even better: “the law,” meaning binding censorship obligations.134 The Hate Speech and Disinformation Codes were the first step in granting the European Commission a say in platform content moderation. The DSA amounted to a complete takeover. [IMAGE WITH THIS CAPTION] An internal read-out from a Google employee describes how a top DSA enforcer stated that the Hate Speech and Disinformation Codes were a precursor to the DSA intended to inflict “reputational damage on platforms.”
The European Commission successfully pressured major social media platforms to change their global content moderation rules, directly infringing on American online speech in the United States. Most major social media or video sharing platforms are based in the United States16 and have a single, global set of rules governing what content can or cannot be posted on the site.17 These rules set the boundary for what discourse is allowed in the modern town square, making them a key pressure point for regulators seeking narrative control to tighten their grip on political power. Critically, platform content moderation rules are—and effectively must be—global in scope.18 Country-by-country content moderation is a significant privacy threat, requiring platforms to know and store each user’s specific location every time he or she logs on.19 In an age where users can freely use virtual private networks (VPNs) to simulate their location and protect their personal information, country-by-country content moderation is also ineffective20—in addition to creating immense costs for platforms of all sizes.21 The internet is global, and platforms govern themselves accordingly. That means that when European regulators pressure social media companies to change their content moderation rules, it affects what Americans can say and see online in the United States. European censorship laws affecting content moderation rules are therefore a direct threat to U.S. free speech.
The European Commission’s focus on global content moderation rules remains: in May 2025, the European Commission explicitly told platforms at a closed-door “DSA Workshop” that “continuous review of [global] community guidelines” was a best practice for compliance with the DSA.31
One European jawboning example is the Hate Speech Code of Conduct, an agreement between the European Commission and four major platforms—YouTube, Twitter, Facebook, and Microsoft.29 The code requires platforms to put in place “Rules or Community Guidelines” that prohibit “the promotion of incitement to violence and hateful conduct.” These rules theoretically track existing European law, which would require platforms to take down hate speech upon notification even without this separate commitment. But the code casts platforms’ removal decisions as the product not of law or state power but of private agreements: platforms’ contracts with users under their Terms of Service, and platforms’ agreements with the Commission under the code itself. Since platforms are in theory enforcing only their own rules when they take down hate speech, users have no clear means to dispute legal interpretations or raise defenses based on European free-expression guarantees. By putting interpretation of a uniform standard in platforms’ hands, the code also homogenizes and “federalizes” European law, despite substantial variation in member states’ actual laws and the states’ supposed authority to maintain their own rules for free expression.30