Tech Roundup: Bluesky's Privacy Problem, YouTube Updates & More

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

• Romania's constitutional court annuls the result of the first round of voting in the presidential election amid allegations of Russian interference.
• Meta says parental consent and age verification for social media use should happen via app store operators such as Apple and Google, as Australia's Senate passes a social media ban for children under 16.
• X and Threads rival Bluesky warns of an uptick in "harmful content posted to the network" as the platform experiences a huge influx of users (currently at 24.5 million).
• That said, Bluesky's DNS-based approach to verifying accounts has raised concerns that it could be misused by malicious actors to create new domains impersonating their targets or snap up old handles as users change theirs to match a domain of their own.
• In response, the company said it will take a more aggressive approach to removing impersonation and handle-squatting accounts, and that it will require parody, satire, or fan accounts to clearly label themselves in both the display name and bio.
• AI company Anthropic releases a new open-source tool called the Model Context Protocol (MCP) that allows users to connect AI assistants directly to specific datasets to inform their responses or carry out tasks; adds new custom styles feature that gives users more control over the writing output.
• Google adds new Gemini extensions for WhatsApp and Spotify, allowing users to send messages or play music directly from the AI assistant; tests a new Android feature called "Lock screen notification minimalism" that aims to declutter the device lock screen by minimising them to a small row of icons that can be tapped to reveal all notifications.
• Bluesky, which has emphasised it won't user generated content for training its own AI offerings, says it is up to outside organisations to respect user consent after an independent researcher briefly makes available a dataset composed of one million Bluesky posts from 320,762 accounts, including information containing when they were posted and who posted them; notes that "Bluesky is an open and public social network, much like websites on the Internet itself, [and] websites can specify whether they consent to outside companies crawling their data with a robots.txt file, and we're investigating a similar practice here."
• While the platform is likely exploring a setting that allows Bluesky users to specify whether they consent to outside developers using their content in AI training datasets, the development highlights another important aspect in that the very open nature of BlueSky means that it's also susceptible to scenarios where anyone can use its Firehose API to grab user data for their own purposes, including used it as AI fodder. While the original dataset is no longer available, it has had the Streisand effect of springing forth several other larger datasets in its place, including one with 298 million posts.
• If anything, it's yet another reminder that anything that's posted online for public consumption  could be scraped for AI training.
• ByteDance-owned TikTok plans to block users under 18 years of age from using beauty filters and says it is tightening its systems to block users under 13 from the platform.
• OpenAI acquires the domain "chat.com"; suspends access to Sora in response to a group of artists leaking access to the tool in protest of the company's treatment of creative professionals.
• Microsoft clarifies that it does not use customer data from Microsoft 365 consumer and commercial applications to train large language models.
• Google succeeds in overturning a US$ 1.7 billion antitrust penalty handed down by the European Union back in March 2019 despite accepting "most of the Commission's assessments" that the company had used its dominant position to block rival online advertisers; says the Commission failed "to take into account all the relevant circumstances in its assessment of the duration of the contractual clauses that it had found to be unfair."
• X adds the option for user to hide engagement buttons and numbers below each post; comes as a study finds that the platform moves slowly to remove AI-made nudes reported as nonconsensual intimate media, if at all, but is quick if they're reported for copyright violations.
• Meta says the company's AI-driven feed and video recommendations have led to an 8% increase in time spent on Facebook and a 6% increase on Instagram; joins hands with OpenAI and Orange to train AI models on African languages, starting with Wolof and Pulaar, addressing a shortage for Africa's thousands of dialects.
• A new analysis finds over 54% of longer English-language posts on LinkedIn are likely AI-generated, as automated prose becomes popular across the internet.
• Google rolls out huddles in Google Chat to let users instantly start an audio-first meeting; proposes Search tweaks in the EU to comply with the DMA, including removing the map showing hotel locations and results in Germany, Belgium and Estonia
• India's competition watchdog launches a probe into Google's restrictive policies for real-money games on Android following a complaint by gaming app WinZO.
• Meta appears to be developing a Threads feature inspired by Bluesky's "starter packs", with profiles "handpicked by people on Threads" for users to follow.
• Snap files to dismiss U.S. state of New Mexico's lawsuit over alleged teen account recommendations to child predators, saying the case misrepresents the state's investigation; accuses the attorney general of intentionally seeking out and friending such accounts before recommendations were made.
• Indian travel and hospitality aggregator MakeMyTrip agrees to acquire expense management platform Happay from fintech CRED.
• Apple seeks to dismiss U.S. Department of Justice's (DOJ) antitrust lawsuit against it over its alleged monopoly in the smartphone market with the iPhone and the device's locked-down ecosystem.
• Music streamer Spotify says it will no longer allow developers building new third-party apps with its Web API to access several features, such as song and artist recommendations.
• Perplexity debuts a shopping feature offering product recommendations and one-click checkout for Pro users in the U.S. as the company ventures into e-commerce and
• Google introduces a new feature on iOS called Page Annotations that injects links on third-party websites that take users back to Google Search to allow users to "quickly get additional context about people, places or things - without leaving the site they're on."
• DirecTV reaches an agreement to acquire Dish TV, Sling TV and the rest of Dish parent company EchoStar's wider TV business.
• Raspberry Pi, the company that sells tiny, cheap, single-board computers, releases Raspberry Pi AI Camera, an image sensor module comes with on-board AI processing and is costs US$ 70.
• Apple marks the final iPod nano and iPod shuffle models to its obsolete products list worldwide, more than seven years after they were discontinued in July 2017; opens an applied research laboratory in Shenzhen, China.
• The U.S. state of California passes the California Neurorights Act into law that treats "neural data" as personal sensitive information, aiming to protect brain data from misuse.
• SpaceX's Starlink satellite internet network crosses 4 million customers, after it started beta services in October 2020 and hit 1 million subscribers in December 2022, 2 million in September 2023, and 3 million in May 2024.
• South Korean electronics giant Samsung rolls out One UI, based on Tizen 8.0, to its smart TVs that were released in 2023.
• Google's YouTube launches granular playback speed controls, an improved miniplayer on iOS and Android, playlist updates that add the ability to create custom thumbnails using AI and a sleep timer.
• The European Union unveils a new Appeals Centre Europe (ACE) as a an Out-of-Court Dispute Settlement (ODS) body for residents to settle content moderation disputes with Facebook, TikTok and YouTube.
• India issues a notice to Wikipedia over bias concerns, questioning if it should be classified as a publisher, after judges called its open editing "dangerous."
• Google says more than a quarter of all new code at the company is generated by AI, then reviewed and accepted by engineers.
• A new study shows that OpenAI ChatGPT routinely mixes up reports from news outlets, providing either entirely incorrect attribution details or fabricating source material when it's unable to locate the correct source for a quote in some cases, potentially risk causing reputational damage to publishers.
• OnlyFans, a popular subscription-based website known for its sexually explicit content, quietly become accessible in China after previously being blocked by government censors.