Tech Roundup: E.U. AI Act, X Grok-1 Model Open Release & More

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

• The E.U. Parliament approves landmark AI Act that puts safeguards on general purpose artificial intelligence, limits on the use of biometric identification systems by law enforcement, bans "AI applications that threaten citizens’ rights, including biometric categorisation systems based on sensitive characteristics and untargeted scraping of facial images from the internet or CCTV footage to create facial recognition databases."
• A new investigation from the European Data Protection Supervisor (EDPS) finds that the European Commission "infringed several key data protection rules when using Microsoft 365."
• Bluesky open-sources its collaborative moderation tool Ozone and plans to let users run their own independent moderation services.
• Discord unveils plans to launch an Embedded App SDK that aims to let developers "build new games and experiences that can be played directly on the platform."
• Apple debuts App Store updates for the European Union to allow qualifying developers to offer apps on their websites via a Web Distribution tool from spring 2024; limits it to developers who had an app installed by over 1 million users the previous year.
• Web browser makers Google, Apple, Mozilla and Microsoft announce Speedometer 3.0 to "measure web application responsiveness by simulating user interactions on real web pages."
• ByteDance-owned TikTok is reportedly planning a separate TikTok Photos app for Android and iOS as it continues to take on Instagram.
• AI company Midjourney bans competitor Stability AI from using its AI image-generating services, after it suffered an outage due to a Stability AI employee allegedly "trying to grab all the prompt and image pairs."
• Amazon rolls out a generative AI feature to help create high-quality product detail pages using information from a seller's external website.
• Anthropic releases Claude 3 Haiku that's tailored for high-volume, latency-sensitive applications, describing it as "the fastest and most affordable model" in its intelligence class.
• Google's YouTube revamps its TV app to make the viewing experience more interactive and giving greater prominence to chapters, shopping, comments and video descriptions; brings YouTube Premium to nine new markets, including Azerbaijan, Kazakhstan, Libya, Morocco, Réunion, Tanzania, Uganda, Yemen and Zimbabwe.
• Microsoft plans to release a unified Teams app on Windows 11 later this year, letting users switch between multiple tenants and personal or work account types; to make Copilot for Security generally available starting April 1, 2024, and expands Copilot Pro to 222 countries with a one-month free trial.
• OpenAI announces partnerships with Le Monde and Prisa Media to bring French and Spanish language news content to ChatGPT and help train its AI models, adding to its existing licensing deals with Shutterstock, The Associated Press and Axel Springer and amid its legal battle with The New York Times over allegedly bypassing paywalls to access entire articles to train its AI models.
• Google DeepMind unveils a generalist AI agent called Scalable Instructable Multiworld Agent (SIMA) that can follow natural-language instructions to carry out tasks in a variety of video game settings an AI agent and play like a human.
• Music streamer Spotify takes on YouTube by adding support for music videos to its mobile and desktop apps in beta in 11 markets, including Brazil, Italy, and the U.K., for Premium subscribers.
• Google follows Apple's footsteps with a new "external offers program" that allows third-party developers to promote offers for in-app features and services in the E.U. region for a fee.
• Meta-owned Instagram internally pilots a Spins feature for Reels that allows other people to swap out text or audio in a user's Reel; to shut down CrowdTangle in August 2024 and replace it with Meta Content Library that's available only for academic and nonprofit researchers.
• Snapchat begins testing Infinite Retention Mode, which lets users save DMs indefinitely, bringing it closer to a pure-play messaging app like iMessage or WhatsApp.
• The U.S. Central Intelligence Agency (CIA) launched a clandestine campaign on Chinese social media aimed at turning public opinion in China against its government, according to a new report from Reuters; promoted allegations that "members of the ruling Communist Party were hiding ill-gotten money overseas and slammed as corrupt and wasteful China's Belt and Road Initiative, which provides financing for infrastructure projects in the developing world."
• Apple agrees to pay US$ 490 million to settle a class-action lawsuit alleging that the company's CEO Tim Cook defrauded shareholders by concealing falling demand for iPhones in China in 2018; to face a lawsuit accusing the company of negligence over the potential stalking risks created by AirTags.
• The Indian government revises its AI advisory to drop plans to require government approval before new AI models are launched after receiving criticism from entrepreneurs and investors.
• The U.S. Federal Trade Commission launches an inquiry into Reddit's licensing of user data to AI companies; Reddit says the FTC is "conducting a non-public inquiry focused on our sale, licensing, or sharing of user-generated content with third parties to train AI models."
• Adult entertainment platform Pornhub disables access to its site in the U.S. state of Texas following a court ruling that upheld a state law (HB 1181) requiring age-verification systems on porn websites; Pornhub calls the law "ineffective, haphazard and dangerous." (Pornhub has already pulled out of multiple states in response to a wave of age verification laws sweeping the country, including in Montana, Utah, Virginia and others.)
• Italian authorities fine TikTok €10 million for fueling the spread of videos likely to harm the "psycho physical safety" of minor and vulnerable users; says the content is "disseminated through a 'recommendation system' based on algorithmic user profiling, which constantly selects which videos to target to each user in the 'For You' and 'Followed' sections, with the aim of increasing user interactions and time spent on the platform so to boost advertising revenue."
• Microsoft-owned LinkedIn confirms it is working on adding puzzle-based games; to also rank companies in the games based on the scores of employees.
• X formally open-sources the AI model underpinning its generative AI chatbot Grok, a 314 billion parameter Mixture-of-Experts model trained from scratch called Grok-1, releasing the "base model weights and network architecture."
• Apple researchers detail MM1, a series of multimodal LLMs with up to 30 billion parameters they say achieve state-of-the-art performance across multiple AI benchmarks.