Tech Roundup: E.U. AI Act, U.K. Microsoft-OpenAI Antitrust Concerns & More

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

• The European Parliament reaches a "provisional agreement" with the Council on landmark rules govering the use of AI in the region (called AI Act); limits law enforcement use of biometric identification systems that use sensitive information and enforces new guardrails for AI systems, requiring companies to "conduct model evaluations, assess and mitigate systemic risks, conduct adversarial testing, report to the Commission on serious incidents, ensure cybersecurity and report on their energy efficiency."
• The U.K. Competition and Markets Authority (CMA) seeks comments on "whether the partnership between Microsoft and OpenAI, including recent developments, has resulted in a relevant merger situation and, if so, the impact that the merger could have on competition in the U.K"; comes as the U.S. Federal Trade Commission is reportedly examining the nature of Microsoft's investment in ChatGPT maker OpenAI, and whether it may violate antitrust laws.
• Meta's WhatsApp launches support for ephemeral voice messages and updates its iOS app to let users send photos and videos in their original quality files and start voice chats without ringing in large groups; tests sharing status updates to Instagram and officially rolls out end-to-end encryption to Messenger by default for one-on-one chats and calls.
• Global smartwatch shipments grow 9% YoY in Q3 2023, with Apple boasting shipments up 7% YoY, Samsung fell 19% YoY, and Huawei grew 56% YoY
• Meta says it doesn't plan to add chronological search results to Threads stating it "would create a substantial safety loophole" and that "spammers and other bad actors pummel the view with content by simply adding the relevant words or tags"; launches tags on Threads, offering support for multiple words and spaces, and no # symbol. (In an interesting twist, posts can only have one tag.)
• Meta announces plans to discontinue the ability for users to chat with Facebook accounts on Instagram beginning in mid-December 2023. (It's not clear what prompted the move, given that the company had previously announced plans to turn on end-to-end encryption across Instagram and Messenger chats in order to make it interoperable with WhatsApp. But it also coincides with the European Commission's decision to regulate Messenger as a "core platform service" under the Digital Markets Act, forcing it to be interoperable with other messaging services. Meta has since sought to push back against the designation, arguing Messenger is a feature of Facebook rather than a standalone messaging platform.)
• ByteDance-owned TikTok expands Ticketmaster integration expands to users outside the U.S., including Australia, Europe and North America.
• Microsoft plans to offer three additional years of monthly Windows 10 security updates, paid for annually, after the operating system's end-of-support date on October 14, 2025; pilots new changes to its Phone Link app for Android that allows users to provide a video stream to Windows PCs, effectively acting as a webcam.
• Discord overhauls its mobile app, offering separate messages and server tabs, new messaging features, a Midnight theme, and speed and data usage improvements.
• Apple puts a stop to Beeper Mini for Android, which let users send iMessages using end-to-end encryption (E2EE), without having to use a new number or log in with an Apple ID, for US $2/month; says it "took steps to protect our users by blocking techniques that exploit fake credentials in order to gain access to iMessage" and that "these techniques posed significant risks to user security and privacy, including the potential for metadata exposure and enabling unwanted messages, spam, and phishing attacks."
• Amazon revamps its Alexa app to improve user experience, making smart home management more intuitive and accessible.
• Microsoft debuts Seeing AI app for Android, more than six years after it was launched in iOS, allowing blind and low vision people to assist with "daily tasks such as reading mail, identifying everyday products, hearing descriptions of photos, and much more."
• Google tests new Smart Search feature in its Files app for Android to help find text, location and objects in images and documents, and Artist, album and title from audio and video files; posts instructions for restoring files in Google Drive for desktop, after users reported last week that some of their files disappeared from the service.
• Apple quietly releases MLX, an array framework to train and deploy machine learning models on Apple silicon.
• Meta rolls out 28 AI characters across WhatsApp, Messenger and Instagram in the U.S.; adds new generative AI features for Facebook posts, discover new experiences with Reels in Meta AI, create and share AI-generated images, and expands its Imagine text-to-image generation feature to all users in the U.S. (Meta says it used 1.1 billion publicly visible Facebook and Instagram images to train Emu, the AI model that powers Imagine.)
• Google debuts Gemini, a multimodal AI model capable of understanding different types of information including text, code, audio, image and video; integrates it to Bard and Google Pixel 8 Pro. (Google has also released a slew of new features for its Pixel devices, including Video Boost for improved video quality by applying computational photography on the cloud and Repair Mode, which locks users' personal data when sending their phones for repair.)
• U.K. regulator Ofcom lays out draft guidance for implementing age checks to stop children accessing online porn services, including "photo ID matching, facial age estimation and credit card checks."
• Amazon tells customers that the company plans to stop accepting Venmo as a payment method on January 10, 2024, after rolling out Venmo support in October 2022; tests a US$ 9.99/month grocery subscription service for Prime members, offering unlimited Whole Foods and Amazon Fresh grocery delivery on orders over US $35.
• Meta unveils Purple Llama, an umbrella project featuring open trust and safety tools and evaluations to help developers build responsibly with AI models.
• Swiss privacy-focused company Proton debuts new photo backup feature for Android that automatically syncs users' photos from their phones to the cloud.
• X begins officially rolling out Grok, its answer to OpenAI's ChatGPT and Google Bard, for Premium Plus subscribers in the U.S.
• Google launches NotebookLM, an "AI-first notebook" demoed at I/O 2023 powered by Gemini Pro, in the U.S.; releases new migration tool to help users move their podcasts library from its standalone podcasts app, Google Podcasts, to YouTube Music ahead of the former's shutdown in April 2024.
• Anthropic finds adding pleas to a prompt that tell its Claude 2 model not to be biased could reduce discrimination based on race and gender.
• A pair of reports from The Washington Post and The New York Times allege that complaints from some senior OpenAI leaders that Sam Altman had been psychologically abusive and that he "was trying to pit [the board members] against each other" were a major factor in the board's decision to fire him.