News insights

Italy’s Data Regulator Again Says OpenAI Violates EU Privacy Law

Report from the BBC

In Brief – Italy’s data protection authority has informed OpenAI that it believes ChatGPT violates several provisions of the EU’s landmark General Data Protection Regulation (GDPR). This is the regulator’s second run-in with the AI phenom, having blocked the service in Italy last April based on similar privacy law concerns. That time, the company resolved the standoff by agreeing to offer a tool to verify the age of Italian users at sign-up and to allow EU users to object to its use of their personal data to train its models. The regulator announced a further “fact-finding activity” at the time, which it now says has found further violations, including challenging the legal basis for the mass collection of users’ data to train the service and concerns that younger users may be exposed to inappropriate chatbot content.

Context – Although the Italian regulator’s specific findings have not been released, the big news is any challenge to the legal basis for “personal data” being used to train AI large language models (LLMs). If information about individual people cannot be part of bulk training material without some form of consent that would fundamentally change the evolution of AI LLMs in Europe. Back in April OpenAI appears to have addressed one aspect of the consent issue by allowing users to block ChatGPT from using their queries and interactions with the service to better train the system. However, that’s a completely different matter than the massive bulk of information collected off the internet that was used, and is still used, to train ChatGPT. If everything online linked to individuals required some manner of consent, the problem would dwarf the issue related to copyright material in training. The EU effort to wrap up their landmark AI Act regulation includes a compromise regime for Foundation Models like ChatGPT. Are data protection regulators rewriting the deal? On protecting young users, ChatGPT recently announced a partnership with Common Sense Media on kid-friendly AI. However, we should know from AI “hallucinations”, when chatbots go off the board and make stuff up, that developers cannot fully contain their output.

Amazon Abandons iRobot Deal in Prologue to Digital Markets Act Fight

Report from the Washington Post

In Brief – Amazon has announced that it will abandon its $1.4 billion plan to buy iRobot, the maker of Roomba robotic vacuums, due to opposition from the European Commission. The EU’s Statement of Objections laid out concerns that Amazon could undermine robot vacuum rivals who sell on the massive Amazon marketplace. The regulator wanted Amazon to guarantee fair treatment for other sellers. The two sides could not agree on how that would work. Amazon will pay iRobot a $94 million breakup fee and iRobot said it will cut 350 employees, 31 percent of its work force, in a bid to return to profitability.

Context – This result is more about Amazon’s DMA compliance than robot vacuums or tech acquisitions. Yes, the EU, who threw the lifeline to Microsoft to save its $76 billion Activision deal, made the kill shot on little iRobot. In both, the UK CMA made the opposite call. They blocked the mega deal and then reversed course after the EU approved. Sure, Microsoft tweaked their concession package. With iRobot, the CMA approved based largely on their analysis that Amazon had little financial incentive to undermine vacuum competitors because they make more money from third-party sales that use their logistics services and advertise. Substantively, the CMA was right, for what that matters. The FTC ended up being a secondary player in both. They lost in court on Microsoft-Activision and were simply slow on Amazon-iRobot. The FTC leadership probably doesn’t mind avoiding a federal court this time. It was not a great case. Finally, back to the EU. Why did iRobot go down and Activision prevail? First, Microsoft was willing to deal. They considered their mega-purchase essential. Amazon clearly did not. The markets were never sold and some suspect buyer’s remorse. Most of all, Amazon appears to have been unwilling to make formal non-discrimination concessions that they must have suspected the EU regulators would try to apply across their full Marketplace under the DMA. So, consider this an announcement from your captain to expect some turbulence from Amazon on the DMA, a la Apple’s plan that has app developers screeching.

Apple’s Announces DMA Policies to Replicate Some “Walled Garden” Security

Report from the New York Times

In Brief – Apple has announced policy changes it says meet the EU Digital Markets Act’s mandates and offer some protection to iPhone users while insisting that the new law lowers security standards. The company will allow app “sideloading” for the first time but will require app developers to meet security standards like those enforced by its App Stores. Apple will institute new systems to approve and monitor sideloaded apps. The plans include major fee changes, including a new 50-euro cents charge for every app download after an app is downloaded one million times in a 12-month period, and a reduction in Apple’s general commission from 27% to 17% (and 15% to 10% for subscription renewals). Like they are doing in the Netherlands, South Korea, and the US, Apple will charge 3% more when an app processes in-app payments using Apple’s payments service. Vocal Apple critics like Epic Games and Spotify harshly criticized the plan.

Context – The DMA challenges Apple’s core user proposition more than the other “gatekeepers“. Most of the alleged anti-competitive or anti-consumer behaviors of the others are non-transparent, hidden inside highly complex operations, or raise questions whether users really understand and consent to how the services work. A core iPhone value proposition is its “walled garden”. Apple has not been surreptitious. They advertise active intermediation to their users, promoting greater privacy, security, and more controlled user experiences. They compare it to the alternative offered by Android. Fundamentally changing the iPhone ecosystem is likely to be more noticeable to users than behind the scenes, black box-type changes to Google search rankings, Meta ad content, or the Amazon Buy Box or how it judges logistics services. Apple is trying to set boundaries around DMA mandates against self-preferencing and requiring interoperability. Apple’s app industry antagonists are already crying foul, and top EU officials are reacting. It will be very interesting to see how, and how quickly, nebulous European Commission DMA enforcement processes work.

FTC and DoJ Fighting Over Review of the Microsoft-OpenAI Relationship

Report from Politico

In Brief – The Department of Justice (DoJ) and Federal Trade Commission (FTC) are in discussions over which agency can probe the antitrust implications of Microsoft’s investments in, and relationship with, OpenAI, the creator of ChatGPT. The FTC initiated talks with the DOJ months ago, but neither is ready to relinquish jurisdiction and allow the other to investigate the details of the tech partnership. The unusual relationship was highlighted by Sam Altman jumping to Microsoft right after he was deposed as CEO of OpenAI. FTC Chair Lina Khan has said that AI has the potential to “supercharge” consumer frauds and anticompetitive practices, sent a lengthy document request and 20 pages of questions to OpenAI in July intimating that the company may have violated federal laws, and has now opened a “study” of investments by tech giants in AI companies.

Context – Even in normal times with normal issues the unusual dual-regulator setup for US Federal antitrust enforcement sees the DoJ’s Antitrust Division and FTC’s Competition Bureau vie for the right to lead specific federal government merger reviews and other antitrust activities. Guessing the winner is a bit of a parlor game for insiders. Back in 2022, when Microsoft’s bid to acquire videogame giant Activision-Blizzard was on the table, both agencies wanted to lead the review, with the FTC claiming expertise and history with the videogame industry while the DoJ was traditionally the lead enforcer for Microsoft. The FTC won that tug-of-war. This time, government officials the world over see AI being of existential importance, or at least the biggest thing since sliced bread, and they must believe that issues related to AI will be front and center for the foreseeable future. So, no surprise that both agencies want to win this prize. In the meantime, the European Commission Competition Authority and the UK Competition and Markets Authority have already each announced that they are looking into whether the Microsoft-OpenAI business partnership is more appropriately considered an acquisition by Microsoft, giving it a level of control that must pass muster with relevant antitrust officials.

Meta to Allow European Users to De-Link Services for DMA Compliance

Report from The Verge

In Brief – Meta is continuing to announce major policy changes in Europe driven by regulatory requirements, this time allowing users to de-link their accounts on Instagram, Facebook, Facebook Messenger, Facebook Marketplace, and Facebook Gaming to comply with the Digital Markets Act (DMA). The move applies to users in the EU plus Iceland, Lichtenstein, Norway, and Switzerland. When users choose to de-link any or all their accounts, the company will no longer use data from one service to influence the others, which Meta says facilitates benefits such as the ability to target ads to a person’s interests, better personalize content recommendations, and share posts between services. Facebook Messenger will be able to be used as a stand-alone service without a Facebook account, while Facebook Marketplace and Facebook Gaming still require a Facebook account but could be used without data sharing with that account. Meta says that using Marketplace without connecting to your Facebook profile will mean communication with buyers and sellers will need to be done over email rather than Facebook Messenger, while Facebook Gaming users without data sharing will be limited to single-player games.

Context – The DMA is imposing unprecedented change on 22 platforms from the initial “gatekeepers” — Amazon, Apple, Google, Meta, Microsoft, and ByteDance. March 6 is a key deadline requiring them to implement the changes they believe are needed to comply with the law’s 18 policy mandates. Recent announcements include Google also letting users turn off data sharing between their major platforms, giving users a choice for their default browser and search apps, and a big change in how Google Search will rank and display so-called “vertical search” services, while Apple will allow side-loading of apps, a massive change in their traditional “walled garden”, and competitors to use “Tap-to-Pay”. Meta de-linking accounts comes on the heels of their roll-out late last year of paid, ad-free subscription options for Instagram and Facebook. All the major changes are currently being applied by the companies only to users in Europe.

Media Representatives Testify at Senate Hearing on AI and Journalism

Report from Ars Technica

In Brief – Witnesses representing media companies and local newspapers testified before the Senate Judiciary Subcommittee on Technology and warned that Artificial Intelligence services will likely harm traditional journalism as they believe large internet platforms have. They called for Congress to require AI companies to compensate media companies when AI systems are “trained” on their copyrighted material. Committee members, including vocal Big Tech critics Sen. Richard Blumenthal (D-CT) and Sen. Josh Hawley (R-MO), also called for new AI regulation, and explicitly exempting AI companies from Section 230 of the CDA. There was widespread agreement that some manner of licensing payments should be required for AI training. Sen. Amy Klobuchar (D-MN), who chairs the Committee’s Competition Subcommittee, asked the witnesses if they also supported her Journalism Competition and Preservation Act that aims to force Google and Meta to pay more money to media companies when their content appears on the giant platforms. The media witnesses responded that it would provide them important relief potentially sooner than an AI licensing regime as the measure has been under consideration for a few years.

Context – As traditional media companies push any angle for government to extract mandatory payments out of tech companies, the most interesting developments continue to involve Meta. It is becoming increasingly clear that the company thinks “news” content often brings more headaches than it is worth. In jurisdictions like Canada where a new law forces them to pay licensing fees when users post news, they have cut off news posting and apparently are learning it does not much reduce engagement. Due to news-inspired controversies and bickering, Meta is reportedly now diminishing their circulation of media content, which media companies have called exploitative in the past but are now worried about the loss of Meta’s free distribution platforms. Backers of bills in Congress and Sacramento on forced news media payments must be watching Meta. Google must be as well, as it has left them as the only wallet left in Canada.

FTC Racks Up Settlements in Campaign Against Location-Based Data Businesses

Report from the Washington Post

In Brief – As part of its campaign to protect consumers from what privacy advocates and its Democratic commissioners call “commercial surveillance”, in particular the collection and use of detailed physical location data gathered by individuals’ smart phones, the Federal Trade Commission (FTC) has reached an agreement with data aggregation company InMarket Media to settle allegations that the company failed to get users’ consent before using their physical location data for marketing and advertising purposes. The deal will bar the company from selling, licensing, or sharing any product that targets users based on sensitive location data. The InMarket settlement follows a separate agreement reached with another so-called location data broker, Outlogic, where the company agreed to not sell location data that could reveal a person’s medical visits, at the time a first by the privacy regulator. In its announcement, the FTC said that InMarket maintained lists of users it described by labels such as “parents of preschoolers” and “Christian church goers”.

Context – In August 2022, the FTC took its first action against a location data aggregator, filing a complaint against Kochava, a major geolocation data broker that the agency argued collected such sensitive location data that it amounted to consumer harm. The case was brought weeks after the US Supreme Court overturned the Roe v Wade decision on abortion access and President Biden called on several federal agencies to take action protecting access to reproductive and abortion services. The FTC claimed that Kochava enabled tracking of individuals visiting such sites, which could threaten their privacy and security. The company argued that it fully complied with federal laws and rules. Last August, Federal District Court Judge Lynn Winmill sided with the company and dismissed the agency’s lawsuit arguing that it failed to show evidence that the sale or use of the data caused any actual substantial injuries, instead describing the agency’s claims as theoretical. The FTC has filed an amended complaint.

As Expected, the EU Will Block Amazon’s Deal to Buy Roomba-Making iRobot

Report from Reuters

In Brief – As fully expected after Amazon did not offer the EU Competition Authority a set of formal concessions to address regulator concerns with their $1.4 billion plan to buy iRobot, the maker of Roomba robotic vacuums, multiple reports indicate that the regulator will reject the deal. The final procedural hurdle is a formal vote of the 27 European Commissioners. News of the EU’s plans have resulted in the shares of the vacuum maker falling a further 30 percent. EU officials began a formal investigation of the acquisition in July and followed up with a Statement of Objections in late November. They laid out several specific concerns on strategies they believe Amazon could pursue to undermine robot vacuum rivals who sell on the massive Amazon marketplace. The Commission wants Amazon to make specific commitments detailing how they will guarantee fair treatment for other vacuum manufacturers.

Context – The Amazon-iRobot saga is clearly about more than the deal. Nobody thinks that the acquisition is that important to Amazon’s business aspirations. Instead, the standoff appears to be a proxy battle over implementation of the Digital Markets Act (DMA), as well as providing insights on current state of EU acquisition policy for Big Tech. First, the DMA. As tech regulation goes, it is a mega event. For the six “gatekeepers” it orders how they operate their 22 biggest platforms. These top platforms are immensely vast and complex, including the behemoth of the Amazon marketplace and its interlocking logistics business. While the UK CMA cleared the iRobot deal and argued that third-party sales, such as competitor vacuums, were likely more profitable to Amazon than in-house vacuums and therefore unlikely to be undermined by Amazon, the Commission appears to want Amazon to be agreeing to specific DMA-compliant behaviors now. And Amazon appears to be passing on that opportunity and waiting until they must. Secondly, the EU rejecting this deal on the heels of it rejecting Booking’s bid to buy eTraveli indicates that very strong concessions will be needed for any acquisitions by the biggest digital platforms.

Apple to Offer Alternate DMA-Compliant App Store and iOS in Europe

Report from GB News

In Brief – As Apple prepares to comply with the EU’s landmark Digital Markets Act (DMA), reports indicate that the iPhone giant will roll out separate versions of its iOS operating system and App Store to allow the “sideloading” of apps for European users. The DMA, which comes fully into effect on March 7, regulates the 22 largest “core platforms” of six digital “gatekeepers”, imposing 18 so-called “Do’s and Don’ts”. Apple has three such platforms – the iOS operating system, Apple App Store, and Safari browser. Among the mandates imposed on the digital giants are several non-discrimination and interoperability requirements that directly target Apple’s “walled garden” iPhone ecosystem where the company controls applications by requiring app developers to use of its App Store and comply with App Store rules, and technically blocks users from downloading apps elsewhere. Apple is also expected to allow developers to add support for third-party payment providers and promote offers to buy services outside the App Store. Despite repeatedly arguing against the legislation’s direct attack on its tightly controlled ecosystem, which it defends as offering users more security and privacy, and a more consistent user experience, a set of fundamental changes to its iPhone system appears set for European users.

Context – Apple’s iPhone may be the platform facing the biggest, or at least most noticeable, DMA changes. Its highly controlled ecosystem has stood in stark contrast to Google’s more open Android where, for example, sideloading has been possible if not facilitated or encouraged. Change is coming to the big platforms, whether through court decisions in the US, or new laws in the EU. The DMA is the biggest driver. One thing we are seeing as March approaches is that the platform giants are not rolling out changes made for the DMA into other jurisdictions. They appear set on offering different services in different geographies. It will be interesting to see how that evolves, including when there are cross-jurisdiction use cases, and how different the services look and feel if the branches extend out years or decades.

OpenAI Rolling Out Policies on Political Campaigns and Elections

Report from CNN

In Brief – OpenAI has announced a set of policies that it claims will reduce the threat of its technology tools being used to undermine election integrity. The company is prohibiting political campaigns from building specialized applications for lobbying or campaigning, as well as prohibiting the creation of specialized chatbots that impersonate political candidates, government officials, or agencies. To reduce the threat of its DALL-E 3 image generating system from creating “deep fakes”, it will add digital credentials to images that indicate their AI origin. Finally, the company is attempting to integrate real-time news, such as through a partnership with Axel Springer, and voting information, including from the National Association of Secretaries of State, with the output of its ChatGPT service when asked for election news and information.

Context – OpenAI is following in the path of the largest social media companies by rolling out a combination of rules, practices, and algorithms to address problematic content. The new political prohibitions don’t aim to block campaign staff from asking ChatGPT for help in writing speeches and talking points. Instead, they aim to limit the ability to build specialized AI-powered political tools on top of the core ChatGPT model. Those are more like business customers than individual users, and the number of such specialized applications being built is probably not so large that a moderately sized team could police their development, root out violators, or shut them down when they are found, such as OpenAI shutting down a bot that claimed to offer responses based on the views of Democratic Presidential candidate Dean Phillips. Then there are “guardrails”, where the chatbot operator instructs the system to override its normal operations and do, or not do, certain things based on key words and issues. Those operate like social media content moderation systems. OpenAI’s plan to add news content or election information along with, or even above, the chatbot’s own creations, fits that model. Finally, there are questions regarding whether the outputs from ChatGPT or any AI chatbot are ideologically slanted. But that’s a different topic.

More Google DMA Changes – Vertical Search Firms Vindicated

Report from Reuters

In Brief – Google is continuing to announce changes to its top services in Europe to comply with the landmark Digital Markets Act (DMA), this time describing how its search engine will rank some so-called “vertical” search sites, as well as plans to give users more “choice screens” to set their default browser and search service and ease the shifting of data from Google services to other providers. Google changing how its search engine treats specialized vertical search services, such as for shopping or travel, is particularly noteworthy as downranking links to vertical search sites in general Google search has been a point of contention for well over a decade. Google already provides Android phone buyers with a search engine choice menu as part of the remedies it instituted to resolve the European Commission’s Android antitrust complaint, and it plans to give a similar choice to Chrome users, as well as offering Android users an additional choice to set their browser default. The new changes become effective March 6.

Context – The DMA is imposing unprecedented change on 22 platforms from six “gatekeepers”. Changing how Google Search treats competitor verticals is noteworthy because Google’s treatment (or mistreatment) of vertical search services kicked off the events that brought us the DMA. In 2010, the EU began investigating allegations that Google was abusing its dominant “horizontal” search service to downgrade vertical search providers and preference its own verticals. The result was the 2017 Google Shopping decision and fine that is still the subject of court appeals. That time lag led to calls to replace antitrust enforcement with regulation. And years of Google remedies that were ad-based mechanisms requiring verticals to pay Google for placement on the search results page never quelled criticism from vertical competitors. In the most recent announcement, Google says that some might object to the heightened placement of vertical comparison sites because they will feel downranked. The EU referees are certain to hear endless advice on how Google should tweak search.

EU Asks for Feedback on Apple’s Offer to Settle Tap-to-Pay Antitrust Complaint

Report from TechCrunch

In Brief – Apple has submitted to European Commission antitrust officials a set of commitments to give third party mobile wallet and payment services providers access “near field communications” (NFC) technology on iOS devices to settle an antitrust complaint initiated in 2022. The highlight of the offer is to give payments services access to most NFC components, free of charge, in what Apple calls “Host Card Emulation (HCE) mode” that the iPhone giant says provides the necessary functionality to securely store payments credentials and complete transactions using NFC but will not grant access to a chip Apple calls the “secure element” that it says enhances the security of transactions using Apple Pay. They further pledge to provide third parties with added functionality including being set as a default payment app and access to authentication features such as FaceID, as well as allow HCE mode to work in stores outside the region. The Commission is asking for public feedback, giving 30 days from the offer being published in the EU’s Official Journal.

Context – This EU antitrust complaint is about payments but is NOT part of the massive fight over “in-app payments” hounding Apple and Google. Those are about app developer fees. This is one of many complaints, such as from Spotify, alleging that Apple discriminates against third-party apps that compete with Apple’s own services. In this case, Apple has blocked competitors from using NFC for in-store “Tap to Pay”, reserving it just for Apple Pay, arguing, as they do in nearly all iPhone policy battles, that they are protecting user privacy and security. If the offer is accepted, and the feedback of competitors such as PayPal will be key, it will resolve the issue before the Digital Markets Act comes fully online in March. Apple’s iOS is a covered platform, so the set of requirements to end self-preferencing and allow interoperability would soon kick in regardless. Note that Apple is not opening Tap-to-Pay for non-EEA iPhone users, but the EEA users will be able to use the capability globally.

Google Offering EU Users Options to Not Share Data Between Services

Report from The Verge

In Brief – As Google prepares to comply with the EU’s landmark Digital Markets Act (DMA), the company has announced that users will be given the choice of whether their data gathered on its most popular services will be able to be shared with the other Google services. The DMA, which comes fully into effect on March 7, regulates the 22 largest “core platforms” of six digital “gatekeepers”, imposing 18 so-called “Do’s and Don’ts”. Google has the most such platforms with eight — YouTube, Search, Google Play, Chrome, Google Shopping, Google Maps, Android, and its ad services. One DMA mandate is to not require that personal data from one core platform be combined with personal data from another platform. Google will therefore now allow users in Europe to opt out of data sharing across its covered platforms. The company is reserving the right to use data from one service with another when it is necessary for completing a task, such as making a purchase, stopping fraud, complying with a legal mandate, or protecting against abuse. Google’s DMA mandates go far beyond data sharing rules.

Context – The DMA fundamentally changes the market dynamics for the gatekeepers — Amazon, Apple, Google, Meta, Microsoft, and ByteDance – as well as their users, both businesses and consumers. One major goal was to replace case-by-case antitrust enforcement, which critics of the digital giants argued took far too long to preserve competition, with up front rules enforced by regulators. It is not remotely clear how the regulatory regime will work in practice. The companies and their platforms are massive and technically very complex. The law requires them to change their policies and practices in accord with the law’s requirements and explain their compliance to the European Commission regulators. It looks like a kind of ongoing “remedies phase” in antitrust enforcement cases, with platform rules and operations subject to ongoing critiques by competitors and public advocates, with regulators as the deciders. Some smaller EU businesses are already complaining that they are not being consulted and their voices are not being heard on DMA compliance plans.

Georgia Judge Allows OpenAI Defamation Suit to Proceed

Report from Bloomberg

In Brief – OpenAI has failed to win the dismissal of a defamation lawsuit filed by a conservative radio host in Georgia who alleged that the company’s ChatGPT chatbot made false claims that he embezzled money from a gun-rights organization. Judge Tracie Cason of the Gwinnett County Superior Court denied the company in a one-page order but did not provide a reason. The case will be an early test of the legal implications of how AI-enabled chatbots can produce so-called “hallucinations”, which are realistic sounding claims that can be wholly fabricated by the service and completely unrelated to real events or facts. Mark Walters, the plaintiff, was the subject of a ChatGPT hallucination produced for a fellow radio host who shared the text with Walters claiming that he had been accused of embezzling money from the Second Amendment Foundation. Walters says he has never been accused of embezzlement or worked for the group. OpenAI argued that it should be free of liability because it provides warnings to users that ChatGPT can produce inaccurate information and users should confirm any outputs, as well as that the service did not know the material was inaccurate and did not have malicious intent.

Context – A key learning since the public release of ChatGPT is that generative AI tools are nothing like traditional databases or search engines. They don’t store and return fixed data. Instead, they compile realistic sounding responses to queries by determining which fragments of text best follow other sequences, all based on a complex statistical model that has ingested and processed billions and billions of text examples pulled from all over the internet. They often deliver accurate facts, credible ideas, and responses that seem sensible or creative, but being “correct” is not the goal of the tools, and we’ve all now heard plenty about chatbot “hallucinations”. Realistic-sounding fabrications, such as those that filled a ChatGPT-prepared legal brief that led to two lawyers being sanctioned in US District Court, or fuel chatbot defamation liability concerns, are basically a design feature of the core technology.

Apple-Epic Games Antitrust Case Ends as Supreme Court Rejects Appeals

Report from Bloomberg

In Brief – The Supreme Court has rejected the appeals of both Apple and Epic Games in their three yearlong legal battle. Apple’s “walled garden” business model survived but the iPhone giant needs to change its longtime policy prohibiting app developers from informing customers how they can make purchases outside the App Store ecosystem to reduce fees. Federal Judge Yvonne Gonzalez Rogers’ initial ruling in 2021 was a split decision that largely went Apple’s way, including determinations that Apple did not operate an illegal monopoly and could justify its App Store policies based on user safety and security. However, she ruled that Apple’s “anti-steering” policy blocking developers from telling customers that cheaper payments options were available off Apple violated California’s Unfair Competition Law and she imposed a permanent injunction prohibiting the company from enforcing it with app developers nationwide. The Ninth Circuit Court of Appeals upheld the full District Court decision in 2022. Epic Games appealed to the High Court looking to overturn the main decision while Apple’s appeal was a focused challenge to the judge’s nationwide injunction that it argued was an unconstitutional overreach setting national policy in a case that was not a certified class action lawsuit and overruled a prior class action settlement.

Context – Apple had the better shot at High Court review. The Ninth Circuit panel backed Apple’s justification for their walled garden model, said that consumers are aware of its restrictions, and that Epic did not provide an alternative model that met those needs and compensated Apple. The justices were never likely to wade into a wholesale review of digital antitrust. Apple’s focus on the ongoing issue of federal courts using national injunctions to set policy was more focused and stronger, but we are reminded that most cert petitions are longshots. That said, Apple has been forced to accept other payments methods in other jurisdictions and has rolled out a similar plan for US developers, which still charges commissions but with a 3% deduction. It’s far less a change than is likely coming for them in Europe due to the Digital Markets Act.

Turkey VPN Crackdown Presages Further Online Restrictions as Election Approaches

Report from the Financial Times

In Brief – Turkey’s telecommunications regulator has directed internet service providers in the country to cut off access to 16 popular virtual private network (VPN) services that the government believes are used to circumvent national laws regulating online content and communications. The move is seen as an indication that Turkey is prepared to aggressively restrict online sources of news and information in the run-up to local elections across the country on March 31. X, formerly Twitter, has also announced that they recently complied with a government order to block 15 Tweets, claiming that the government threatened to shut down the platform. Critics complained that banning VPNs is one of the most aggressive moves to enforce online censorship, for example by China, and that while Turkey has taken action against VPNs in the past, the latest move appears more widespread and effective.

Context – Modi’s India, Erdogan’s Turkey, and Putin’s Russia intermittently make news for their latest efforts restricting online speech. And China’s Great Firewall seems to be an aspirational goal for authoritarian governments around the world. But the trendline is up almost everywhere. Vietnam, Indonesia, and Singapore in Asia. Africa is a hotbed, with a 2021 report claiming that 31 of 54 African countries had blocked social media platforms since 2015. But Western democracies are also active at pressing platforms to police a wide range of objectionable online communications. The UK, Germany, France, and Australia have been stalwarts. The EU’s Digital Services Act deserves meaningful censorship scrutiny as evidenced by the fact that EU officials quickly challenged X, Meta, and TikTok to block content that said officials found objectionable related to the Hamas attack on Israel and resulting conflict. The willingness of digital platforms to comply seems to mostly to be related to the size and importance of the national market making demands. In the US, the debate is framed by the First Amendment and how the Supreme Court applies its strictures in the context of online platforms. Two huge cases are coming up this year.

Judge Pauses Ohio Social Media Law Requiring Parental Consent for Teens

Report from Ars Technica

In Brief – A federal judge has issued a temporary restraining order blocking Ohio from enforcing the Social Media Parental Notification Act that requires social media companies to verify whether users are 16 or older and get parental consent for younger users to access services like YouTube, Instagram and TikTok. NetChoice, a trade group that includes a number of platform giants among its members, filed suit to block the law and alleges that it violates the 1st and 14th Amendments. Judge Algenon Marbley focused on 1st Amendment concerns and said that the state would likely not be able to show that it was narrowly tailored enough to meet constitutional scrutiny. The judge highlighted the lack of clarity regarding the services that were even covered by the law, pointing out that the list of 11 criteria was largely undefined and malleable, as well as questioning “an eyebrow-raising exception for ‘established’ and ‘widely recognized’ media outlets whose ‘primary purpose’ is to ‘report news and current events’” without defining how one determines which media outlets qualify and “practically invites arbitrary application of the law.” The next hearing is scheduled for February 7.

Context – Although the data showing causality between teen mental health and social media use is thin, the effort to cordon off parts of the internet and regulate how platforms serve teens is proceeding apace, globally. Congressional legislation has bipartisan backing but has been tripped up by broader partisan division. US States, especially under Republican leadership, have been quick off the mark, but California, France, and the UK are examples of progressive governments trying to create a more regulated version of the internet for teens as well. Like with Ohio’s law, the 1st Amendment is proving a tough hurdle in the US. Major social media design negligence suits may prove more impactful. They aim to circumvent Sec. 230 protections and are proceeding to trial at the federal and state level. Meta’s major changes to content practices for teen users should be seen in this light.

EU Court Advocate Backs Commission’s Google Shopping Antitrust Decision

Report from Reuters

In Brief – A legal adviser to the European Court of Justice (ECJ) has submitted a formal opinion that back’s the European Commission’s 2017 decision that Google violated EU antitrust laws by giving its online shopping recommendation service an illegal advantage over rivals in its search results and fined the company 2.4 billion euros. Widely known as the Google Shopping case, the Commission ruled that the company engaged in unfair self-preferencing on its dominant generalized Google search service by directing users to its specialized Google Shopping service rather than to the sites of comparison-shopping competitors. Google has been engaged in legal appeals since. The General Court in Luxembourg, the EU’s second highest, rejected Google’s appeal in 2021. Opinions from the High Court’s advocate general are not legally binding on the judges but they often form the basis of rulings. A decision from the ECJ is expected in a few months.

Context – The EU Competition Authority began investigating allegations that Google was self-preferencing on its dominant “horizontal” search service, including downgrading “vertical” search competitors, in 2010. They eventually focused on Google Shopping and the comparison-shopping vertical. The 2017 Shopping decision and fine was the first of three antitrust decisions adverse to Google in the EU, followed by the Android decision in 2018 and the AdSense decision in 2019. The Shopping decision and Google’s remedies have never quelled criticism from companies that compete in verticals such as travel, jobs, and local reviews. Many complain that the Commission should have addressed Google conduct in other verticals by now, and that the shopping remedy, to replace free search results with an ad-based system, is unfair to competitors. The length of time involved in the Google antitrust cases were also a driving force behind the EU’s Digital Markets Act (DMA) that aims to proactively regulate “gatekeeper” platforms to achieve competition policy results without multiyear enforcement cases. Remedies to stop self-preferencing on Google search is certain to be a major point of contention with the DMA.

The Maryland Digital Advertising Tax Saga – Back to Step One

Report from AP

In Brief – A three-judge panel of the 4th Circuit Federal Court of Appeals clarified the legal path for businesses engaged in online advertising to challenge the constitutionality of Maryland’s first-in-the-nation digital advertising tax. Trade groups led by the US Chamber of Commerce challenged the law enacted in 2021 on numerous grounds, including that it violated the federal Permanent Internet Tax Freedom Act (PITFA), the Commerce and Due Process Clauses of the Constitution, and that a provision prohibiting businesses paying the tax from including the amount as a separate line-item on customer bills violated the 1st Amendment. The appeals court ruled that the 1st Amendment issue belongs in federal court and ordered the District Court to rule on the question. The appeals court said that the other issues needed to work through the Maryland state tax adjudication system, which includes administrative processes followed by state courts, before any appeals to federal court.

Context – A few US States have dabbled with taxes that replicate the “digital services taxes” enacted by countries like France and India targeting platforms like Google, Meta, and Amazon. Maryland was first to act. By taxing online advertising but not offline advertising, the law appears to clearly violate PITFA, which explicitly prohibits states from taxing commercial activity on the internet that is not taxed the same way offline. However, getting to a ruling has been a morass. A federal judge initially ruled that the federal Tax Injunction Act required that the PITFA and constitutional challenges needed to go to Maryland state court, but the 1st Amendment issue was for federal court. Then a Maryland state court judge dismissed the law for violating PITFA and the US Constitution. Done? The original federal judge then dismissed the 1st Amendment challenge to the law as moot. But the Maryland Supreme Court tossed out the state court dismissal as procedurally premature because the businesses had not first exhausted all administrative remedies. Now the Federal District Court must restart the 1st Amendment challenge to the law and the state institutions will ponder how the law does not violate the clear language of PITFA.

Amazon Lets Deadline Pass for Formal Concessions to the EU on iRobot Deal

Report from the Wall Street Journal

In Brief – Amazon did not submit any formal concessions to the EU Competition Authority to address regulator concerns with their plan to buy iRobot, the maker of Roomba robotic vacuums, before the traditional deadline for concessions in an EU acquisition review passed. That news sent iRobot shares falling the following day. EU officials began a formal acquisition investigation in July and followed up with a Statement of Objections in late November. They laid out specific concerns on strategies they believe Amazon could pursue to undermine robot vacuum rivals who sell on the massive Amazon marketplace. EU Commissioner Didier Reynders said that he wanted Amazon to make commitments to provide fair treatment for the other vacuum manufacturers. The regulators have until February 14 to approve or reject the $1.4 billion deal.

Context – The review of this relatively small acquisition is interesting on a few levels. First, the UK CMA already cleared it. The CMA noted that Amazon’s most profitable ecommerce business line often involves sales by third parties who buy Amazon logistics and on-site advertising. Commissions often reach 50% in those cases, with profit margins far above Amazon’s first-party retail sales. The CMA therefore considered it unlikely that Amazon would engage in a strategy to penalize highly profitable third party sales to boost less profitable first party sales. Third party sales that involve Amazon logistics are so profitable that Amazon recently settled an antitrust investigation with the EU by agreeing to changes that boost the prospects of sellers who don’t buy logistics. Does the Commission suspect that if that agreement is successful that Amazon might shift to preferencing their own goods? Amazon is also now dealing with implementation of the landmark Digital Markets Act (DMA) that raises similar issues of self-preferencing and fair treatment for third parties. Amazon may not want to make concessions to accomplish a relatively small acquisition that becomes a precedent that is applied marketplace-wide by the EU under the DMA processes.

EU Competition Authority Will Investigate Microsoft Relationship with OpenAI

Report from the Financial Times

In Brief – The EU Competition Authority has announced a series of actions aimed at protecting and promoting competition in the markets for “Generative AI” and virtual reality services, including investigating whether Microsoft’s massive investments in OpenAI should be formally reviewed as a merger. The EU’s scrutiny of the Microsoft-OpenAI relationship follows similar action by the UK Competition and Markets Authority (CMA) last month, inviting “all interested parties” to share their views on the nature and implications of the relationship between the two AI giants. Along with reviewing the relationship between the two AI giants, the EU competition regulator has also released a public call for comments from interested stakeholders, including regulatory experts, academia, industry, and consumer organizations, on the competition issues and conditions impacting “Virtual Worlds” and generative AI, with submissions due by March 11.

Context – The unusual part played by Microsoft in last fall’s internal drama at OpenAI has piqued the interest of regulators. When Sam Altman was deposed as CEO by the start-up’s Board of Directors, he reportedly responded by quickly jumping into a top AI development role at Microsoft. Many OpenAI employees threatened to go with him. Altman was soon restored at OpenAI with a reconstituted Board that included a Microsoft representative. Not only are the UK and EU regulators formally looking, the FTC is reportedly studying it as well, and it’s a decent bet that they will make that inquiry more formal. One mantra of tech regulation advocates is that governments should not make the “same mistakes” with AI that they made with social media, apparently meaning regulating more and sooner. On competition policy, they often say AI is quickly becoming dominated by the same digital giants. However, OpenAI is the leader in LLM chatbots, having outperformed giants like Google, Microsoft, Apple, Amazon, and IBM, while Clearview AI pulled off a similar trick beating the giants to an apparently functional facial recognition service.

Texas-Led State AGs v Google Antitrust Trial Set for 2025

Report from Reuters

In Brief – Federal District Judge Sean Jordan has set a March 31, 2025, start date for the antitrust trial pitting a coalition of 16 states led by Texas against Google, who is accused of abusing its dominance in digital advertising services. It will kick off with jury selection. The Texas-led antitrust suit focused on advertising technology was filed in October 2020 in parallel with a US Department of Justice-led antitrust suit focused on Google’s search business. The DoJ case moved to trial much quicker. They engaged in weeks of testimony last fall and will reconvene in May for final arguments. On the other hand, the Texas-led suit was bogged down by legal battling over the proper federal court venue. Texas’s AG wanted the suit fought out in Texas, and eventually prevailed after Congress changed the applicable federal law on venue selection, but it took years.

Context – Google is in the middle of a gauntlet of federal antitrust trials. Their first result went very badly. A jury in California sided comprehensively with Epic Games, ruling that Google violated antitrust law in the operations of its Android app store and in-app payments service. Google decisively lost where Apple largely prevailed two years earlier on the same charges, and this despite the fact that Google’s Android system is clearly far more open to competitors than Apple’s iPhone. Two factors that hurt Google against Epic were the complexity of the Google business model and the fact that the Epic v Google trial was in front of a jury while Epic v Apple was decided by a judge. Apple’s trial also proceeded much quicker, and while Google may have appreciated the delays at the time, they also may have been hurt by jury members living through more years of generic media bashing of “Big Tech”. The antitrust sharks circling Google certainly appear to sense the value of battling the giant in front of juries. The DoJ filed an additional adtech-focused antitrust complaint against Google last January and it has a scheduling hearing this month in federal court in Virginia. Like with the Epic Games trial, and the Texas adtech trial in 2025, the DoJ is asking for a jury trial, possibly as soon as March 2024.

Biden Administration Labor Department Releases New Gig Work Rule

Report from the Washington Post

In Brief – The US Department of Labor (DoL) has released a final rule on employee classification changing the criteria businesses should use when they determine if a worker should be an employee or an independent contractor. The aim is to push businesses to classify more workers as employees, a long-time priority of Organized Labor and President Biden. The new rules establish a “totality-of-the-circumstances” framework that weighs six non-exhaustive factors, including the degree of permanence of the work relationship, the extent to which the work is an integral part of the business, and the use of the worker’s skill and initiative. The new rule replaces one from the Trump DoL that focused on how much control a company has over the worker and a worker’s “entrepreneurial opportunity”. Although “Gig” platforms have been a media focus of the battle, traditional industries like custodial services, home-healthcare, entertainment, and construction have long used contractors. Major business associations quickly expressed opposition to the move and the rule will face legal challenges.

Context – The Biden Campaign is checking off a 2024 box for Organized Labor. And since his DoL took three years, the courts likely won’t resolve the legal questions until after the election. Perfect. If DoL engages in some targeted enforcement, don’t expect big Gig Work platforms, look for something like home care. The high-water mark for worker reclassification in the US was California’s AB 5 enacted in 2019. It was backed by candidate Biden. But California voters exempted Gig driving and delivery platforms at the same time Biden was elected. The issue then bogged down in the Democrat-controlled Congress and even in Democrat-dominated states. Opposition from skilled freelancers was key. This rule confirms that the issue is now a DoL version of Net Neutrality at the FCC. Each party will enact their rules when they take over and court challenges will ensue. Rinse and repeat. The most interesting questions will be whether federal judges apply the “Major Questions Doctrine”, and whether the EU’s Digital Labor Platforms Directive will rise from the mat and be enacted this spring.

The New York Times Sues OpenAI and Microsoft for AI Copyright Infringement

Report from the Wall Street Journal

In Brief – The New York Times (NYT) has sued Microsoft and OpenAI for alleged copyright infringement. The media giant claims the technology companies illegally trained their AI systems on millions of pieces of NYT content without permission and talks to agree on licensing payments failed to reach agreement.

Context – As copyright holders file another AI lawsuit every week, why did the New York Times’ complaint make such news? They are huge and influential, but their arguments are not unique. As everyone following “Generative AI” knows, the biggest chatbot systems were “trained” on almost all the text on the internet, which obviously includes massive amounts of copyrighted material. The first few important rulings from federal judges on the copyright complaints seem to confirm that they grasp that neural networks are not traditional databases. They learn from data rather than store it. One key thing to watch for is systems that routinely produce actual copies. The recent rulings noted that the systems were not creating copies. The NYT lawyers read them and included claims that ChatGPT did make copies, although it’s reported that it took some aggressive maneuvers, including submitting verbatim opening paragraphs of stories as a prompt. OpenAI’s public response explains their view on how users can abuse their system. But the biggest issue remains how courts apply the concept of training as “fair use”, which is a relatively robust exception in US copyright law. And don’t forget the money. The suits want to compel payments. One of the most thoughtful comments regarding payments for training came from AI researcher Chris Collison-Burch at a House hearing last May (see the exchange at 1:13:30). He noted that when every piece of text on the internet trains an AI system, it means that no copyright holder, even giants, accounts for much. The denominator is too big. So payments for training would either be very, very small, or if a decent sum went to everyone, they would become prohibitively expensive for all but the very largest companies. It is the giants negotiating payments with big media companies.

Deal on EU Digital Labor Platforms “Gig Work” Directive Falls Apart

Report from EuroNews

In Brief – The mid-December deal between the European Council and Parliament that appeared to clear the way for the early-2024 enactment of a Digital Labor Platforms Directive to regulate Gig-style work platforms in the EU fell apart just prior to the winter break, derailed by EU member state governments led by France. Spain, which held the rotating position of Council presidency for the second half of 2023, had negotiated an agreement with the Parliament on the contentious issue of worker classification, setting forth five conditions of platform control over work, whereby if two were met then there would be a rebuttable presumption that the workers using the platform were employees. Rather than back the agreement, a majority of member state governments bolted, leading the Spanish to pull the matter from consideration in the last meeting of the Council before the winter break. They publicly handed the Gig work regulation portfolio off to the Belgian Government, which now holds the six-month rotating Council presidency. Renegotiations with the Parliament’s delegation would need to be wrapped up by mid-February if legislation is to be considered by Parliament before the EU-wide June elections.

Context – The EU Commission proposed an initial draft of the directive in December 2021 aiming to standardize when platform workers should be classified as employees, with applicable employee rights and benefits, rather than as independent contractors. Finding a balance between ending so-called “phony” independent contractors while also enabling new work models that give skilled freelancers valuable flexibility and independence has proven very challenging. In the US, legislation at the federal and state level, even in Democrat-dominated states, has also been stymied since California’s largely Democratic voters exempted Gig-drivers from the state’s reclassification law AB 5 in 2020. In the UK, the High Court recently confirmed that the level of platform control is key to worker classification there too, with Deliveroo drivers adjudged to be contractors while Uber drivers are classified as workers.

Google Reaches Settlement in Chrome Incognito Tracking Class Action

Report from the Washington Post

In Brief – Google has agreed to the terms of a settlement with the plaintiffs in a US consumer class action lawsuit arguing that the company deceptively tracked users who selected the supposedly private Incognito mode of their market-leading Chrome browser and therefore believed no tracking was taking place. The terms of the settlement, including how much the company has agreed to pay in damages, and which users will qualify for how much compensation, have not been released, but estimates of the potential damages from the case reached $5 billion. Federal District Court Judge Yvonne Gonzales Rogers will rule on whether the settlement is acceptable by February 24.

Context – Location tracking, both physically in terms of smart phones, and virtually in terms of web browsers, is a ubiquitous aspect of modern digital service and facilitates many of the mobile and web tools that people rely on. However, Google, as well as privacy-champion Apple, have faced legal and regulatory challenges alleging that when they have told users that they could choose not to be tracked, there was often still tracking of various sorts going on. An investigation by AP in 2018 revealed that when Android phone users selected the option to turn off “Location History” between 2014 and 2017, significant tracking still occurred on various apps and for advertising. The company has faced lawsuits and regulatory actions in many markets, including the US, Europe, Australia, and reached a $395 million settlement with a coalition of 40 State AGs. The Incognito mode complaints follow a similar line, namely that claims made to users that they could choose not to be tracked were less than they appeared. Apple is facing a rapidly growing collection of consumer class action lawsuits following reports that the company engages in surreptitious and deceptive data collection, by a range of Apple’s apps, including the App Store, Apple Stocks, Apple Music, and Apple News, even when users have chosen Apple privacy settings where they explicitly request not to have their data collected.

Federal Judge Rejects X’s Effort to Block California Content Moderation Law

Report from Reuters

In Brief – US District Judge William Shubb has rejected X platform’s bid to block a California law mandating that social media platforms submit bi-annual reports to the state’s Attorney General detailing the platform’s content moderation terms of service and how they have been applied to a range of controversial or objectionable content, including hate speech, racism, extremism, radicalization, disinformation, harassment, and foreign political interference. X, formerly known as Twitter, challenged AB 587 last September, largely on First Amendment grounds, and asked the court to issue an injunction before it went into effect on January 1, 2024. Judge Shubb denied X’s motion, ruling that social media terms of service and the reports compelled by the law were commercial speech that warrants less stringent First Amendment protections, referencing the Supreme Court’s 1985 Zauderer decision as well as recent rulings by the Fifth Circuit and Eleventh Circuit Courts of Appeals on the Texas and Florida social media content moderation laws that will be reviewed by the Supreme Court this spring.

Context – Platform policy expert Mark Masnick argues that Shubb’s decision completely misapplies the Zauderer standard, and that AB 587 is clearly intended to influence non-commercial online speech. The law is part of a partisan back-and-forth between Florida, Texas, New York, and California. The Republican-led states acted first, legislating in 2021 against alleged anti-conservative “viewpoint” discrimination by large social media platforms. Democrat-led New York and California followed with laws pushing platforms to better police hate speech and harassment. Florida’s law was largely blocked in the US Eleventh Circuit, although the “transparency” provisions were treated separately by the court, a fact noted by Shubb. Texas’s law was upheld in the Fifth Circuit but later enjoined by the Supreme Court. The High Court has agreed to hear arguments on the First Amendment implications of the Florida and Texas laws this spring, but the mandates on “transparency” rules may or may not be part of their eventual ruling.

Commissioner Margrethe Vestager Meeting with Big Tech CEOs This Week

Report from Reuters

In Brief – Commissioner Margrethe Vestager, who leads the European Commission’s Competition Authority and serves as a top Commission policymaker on digital issues, is in California this week and will be meeting with top executives of some the largest digital and AI industry companies, including Apple, Google, Nvidia, and OpenAI. After a three-month hiatus from her role as the EU’s top competition policy official while she unsuccessfully made a bid to be the next head of the European Investment Bank, Vestager is back in charge of her digital policy portfolio focused on constraining the tech giants through vigorous antitrust enforcement. Her schedule includes in-person meetings with Apple CEO Tim Cook, Google CEO Sundar Pichai, Nvidia CEO Jensen Huang, and OpenAI CTO Mira Murati.

Context – As the EU rolls out the Digital Markets Act (DMA), its landmark competition policy-inspired regulatory scheme for the largest digital “Gatekeepers”, a top open question is how it will operate side-by-side with traditional antitrust enforcement. The European Commission has named Amazon, Apple, Google, Meta, Microsoft, and ByteDance as gatekeepers and will be requiring them to apply 18 policy “Do’s and Don’ts” to a total of 22 of their core platform services, such as the app stores and mobile operating systems of Apple and Google, as well as Google’s advertising platforms, Maps, and Shopping platforms. A number of the platforms now regulated by the DMA, such as Apple’s App Store, Google Shopping, and Google Advertising, already face antitrust enforcement actions by the EU Competition Authority, and Vestager has said traditional antitrust enforcement will continue in some manner in conjunction with the DMA regime. Furthermore, she has been a regular contributor to the debate about regulating artificial intelligence, including through the EU AI Act that might be finalized this spring. Finally, AI-chip giant Nvidia has reported that competition regulators, including in Europe, have inquired about its dealings with the largest cloud services providers.