Tech Roundup: Apple E.U. Scrutiny, TikTok Whee & More

[A recurring feature on the latest in Science & Technology.]

• The European Commission formally accuses Apple of breaching the Digital Markets Act (DMA) through its App Store by preventing "app developers from freely steering consumers to alternative channels for offers and content"; raises concerns about its new contractual terms for third-party app stores.
• eSafety Commissioner, Australia's online safety regulator, waters down new rules to force tech companies to detect child abuse and terror content on encrypted messaging and cloud storage services following concerns of mass surveillance.
• Apple discontinues its "buy now, pay later" (BNPL) service, Apple Pay Later, in the U.S. a little over a year after its launch, as it readies a global installment loan offering when checking out with Apple Pay.
• Meta releases a flurry of new AI models, including Chameleon (which combines text and images as input and output any combination of text and images), Joint Audio and Symbolic Conditioning for Temporally Controlled Text-to-Music Generation (JASCO), and AudioSeal, which it claims is the first audio watermarking technique to detect AI-generated content online.
• Nvidia surpasses Microsoft to become the world's most valuable public company amid growing demand for AI chips used in data centers.
• ByteDance TikTok says it offered the U.S. government in 2022 the power to shut the platform down if it failed to address lawmakers' data protection and national security concerns, as it commences its legal fight against legislation that will ban the app in the country unless it's sold off; argues divestiture is not possible technologically, commercially or legally, and claims the law violates Americans' free speech rights. (The case is expected to be heard in the U.S. Court of Appeals for the District of Columbia in September 2024. The company has also said it has spent obver US$ 2 billion to protect U.S. user data as part of an initiative called Project Texas. The development comes as the U.S. Justice Department has allegedly dropped claims that the misled U.S. consumers about their data security in a forthcoming suit accusing the company of children's privacy violations.)
• Google-owned YouTube tests Notes, an experimental crowdsourced tool to let users add context to videos, similar to X's Community Notes, on mobile in the U.S. in English; begins cancelling YouTube Premium subscriptions obtained for a cheaper price by using a VPN, closing out a loophole that people may have abused to save on subscription costs. (For example, while Premium costs US$ 13.99/month in the U.S., it's just US$ 1.54/month (₹129) in India.)
• Swiss privacy-focused app developer Proton establishes the Proton Foundation to transition to a non-profit foundation model, similar to Signal and Mozilla.
• The U.S. Federal Trade Commission refers to the Justice Department a complaint against ByteDance TikTok for alleged violation of child privacy laws in the country; says its probe "uncovered reason to believe named defendants are violating or are about to violate the law and that a proceeding is in the public interest."
• Adobe announces a tweaked version of its terms of service agreement that makes it clear the company will not train AI on user content stored locally or in the cloud following backlash over changes that initially seemed to suggest it could use people's work to train its AI models; clarifies that work submitted to the Adobe Stock marketplace can be used to train Adobe Firefly.
• Mozilla announces plans to acquire Anonym, which was founded in 2022 by former Meta executives and aims to offer privacy-preserving digital advertising solutions, as it aims to position the technology squarely against that of Apple's (Privacy Preserving Ad Measurement) and Google's (Privacy Sandbox).
• The U.K. government's Department for Transport is using Amazon's Rekognition system to scan the faces of people taking trains as part of a country-wide trial with an aim to "detect people trespassing on tracks, monitor and predict platform overcrowding, identify antisocial behavior ("running, shouting, skateboarding, smoking"), and spot potential bike thieves, according to a report from WIRED.
• The U.K. Competition and Markets Authority (CMA) opens a preliminary probe into Hewlett Packard Enterprise's (HPE) acquisition of Juniper Networks earlier this January to determine if the merger resulted in a "substantial lessening of competition within any market or markets in the United Kingdom for goods or services."
• OpenAI co-founder Ilya Sutskever, who departed the company in May 2024 over AI safety concerns, launches a new company named Safe Superintelligence Inc. (SSI) that aims to singularly focus on safety over short-term commercial pressures.
• Microsoft debuts a new AI model called Florence-2 to handle a variety of computer vision and vision-language tasks; shows off AutoGen Studio, a "low-code interface" for building multi-agent workflows.
• Apple reportedly holds talks with Baidu, Alibaba, Baichuan AI, and others to help offer Apple Intelligence in China to compensate for OpenAI ChatGPT's absence in the country. (How Apple intends to bring its AI features to China will be interesting to watch, especially as Apple's market share is being eroded by local players like Huawei and China has enacted various regulations that, among other things, require companies developing AI models to get approval for their commercial use. This is aside from the fact that Beijing heavily censors internet content, posing challenges for companies that offer generative AI capabilities and how they could respond to questions on sensitive topics in China.)
• Apple's upcoming iOS 18 adds a new API called AccessorySetupKit, letting third-party accessory makers build the same quick pairing experience as Apple accessories like AirPods; to withhold the release of Apple Intelligence, SharePlay screen sharing, and iPhone Mirroring on the Mac in the E.U. this year, citing concerns about DMA's interoperability requirements.
• SpaceX unveils Starlink Mini, a compact version of its satellite internet antenna "that can easily fit in a backpack", for US$ 599.
• Google says it will make user reviews of movies, TV shows, books, albums and video games visible under one public profile page starting June 24 by default (that said, users can opt out of it).
• Meta rolls out the ability for users to restrict their Instagram Live broadcasts to Close Friends, with the option for three other users to join; launches its much-awaited Threads API, letting developers publish posts, fetch content, manage replies, and view analytics, after an API beta in March 2024.
• Anthropic launches Claude 3.5 Sonnet, which beats its flagship model Claude 3 Opus and outperforms rivals on a wide range of evaluations; debuts Artifacts, which lets users see, edit, and build upon content generated by Claude such as code snippets, text documents and website designs. (With all major AI models more or less achieving similar capabilities, it's won't be long before these LLM providers start competing on price and specialised use cases, emphasising on different ways to augment their offerings to stand out.)
• Perplexity courts trouble after it's found that the AI company is ignoring the Robots Exclusion Protocol and using headless browsers to scrape web content that has been explictly marked as forbidden to visit by web crawlers and other web robots, despite claiming to the contrary, a new investigation from WIRED reveals; also finds its answer engine to be prone to hallucinations and inaccurately summarising articles. (Perplexity has since said that it's not ignoring the Robots Exclusion Protocol and that it also relies on third-party web crawlers.)
  
• Password management app 1Password announces the ability for users to regain access to an account when a master password and secret key are lost using recovery codes.
• X plans to make the ability to create live streams a premium feature, making it the only major social platform to charge for the feature.
• Google pauses its tests allowing real-money games on the Play Store in countries like India, Brazil and Mexico; says supporting "real-money gaming apps in markets without a central licensing framework has proven more difficult than expected."
• OpenAI buys enterprise search and analytics startup Rockset for an undisclosed sum, making it the AI company's first acquisition; to "integrate Rockset's technology to power our retrieval infrastructure across products."
• Music streamer Spotify debuts a new Basic streaming plan in the U.S. for US$ 10.99/month, allowing users to experience all the benefits of a Premium plan minus the monthly audiobook listening time; now lets all podcasters upload videos, as it surpasses 250,000 shows.
• Website building platform Squarespace announces plans to sell its restaurant reservation service Tock to American Express for US$ 400 million, the exact amount that Squarespace paid for Tock in March 2021.
• Meta draws the ire of users and photographers after it mislabels real photos -- but retouched or edited using other tools -- "Made with AI" across its products.
• Google launches a Gemini app for Android in India, supporting English and nine other languages (Hindi, Bengali, Gujarati, Kannada, Malayalam, Marathi, Tamil, Telugu and Urdu); expands Gemini integration in Messages to all Android devices with 6GB of RAM or higher.
• AI startup Runway unveils Gen-3 Alpha, a model to generate high-quality, highly realistic 10-second-long video clips.
• Google's DeepMind AI research labs DeepMind outlines its video-to-audio (V2A) technology to take raw pixels of videos and generate synchronized audiovisuals.
• Microsoft adds a new My Phone shortcut to the Windows 11 share window to allow users easily share files to an Android device once it's paired with the PC using Phone Link.
• TikTok quietly pilots Whee, a mobile app for photo sharing with friends and family, in markets like Australia, Pakistan, Thailand and Turkey, more than two months after it began widely rolling out Instagram-rival TikTok Notes; revamps the landing pages for major locations with information related to food and drink, hotels, parks and things to do with an aim to surface relevant videos.
• Open-source AI chat platform LibreChat, which offers integration with AI providers OpenAI, Google, Anthropic others, attempts to differentiate itself by offering users to locally host models in their own systems and run them entirely offline.
• The U.S. state of New York officially passes legislation that requires parental consent for social media companies to use "addictive feeds" powered by recommendation algorithms on kids and teens under 18 (Stop Addictive Feeds Exploitation for Kids act) and limit data collection on minors without consent (New York Child Data Protection act).
• Meta lifts restrictions on election-related queries in its Meta AI chatbot in India and opens it to all users in the country after initial testing, as Google continues to apply these limits out of an abundance of caution.