Google NotebookLM Expands Video Overviews to 80 Languages, Enhances Audio Summaries ↗
Google’s NotebookLM Video Overviews now support 80 languages—including French, German, Spanish and Japanese—up from English only. Audio Overviews also deliver more detailed non-English summaries, with an option for shorter highlights. These global updates roll out over the next week to better serve NotebookLM’s multilingual users.
YouTube tests AI video sharpening on Shorts without opt-out, sparking transparency concerns ↗
YouTube quietly rolled out AI-powered video sharpening and denoising on Shorts, altering creators’ uploads without warning or opt-out. Users spotted odd artifacts and oversmoothing; Google insists it’s “traditional machine learning,” not generative AI. After YouTuber Rhett Shull flagged the changes, Google confirmed the test but hasn’t said if creators can disable it. Unlike Pixel 10’s transparent AI labels for photos, YouTube videos remain unlabeled, raising transparency and creator-reputation concerns as AI edits spread.
Andreessen Horowitz, OpenAI Exec Fund $100M Super PAC to Shape AI Policy in Midterms ↗
Silicon Valley heavyweights like Andreessen Horowitz and OpenAI’s Greg Brockman have funneled over $100 million into a new super-PAC network, “Leading the Future,” to influence next year’s midterms. The group will use campaign donations and digital ads to push for light-touch AI rules, oppose candidates favoring stricter limits and borrow tactics from pro-crypto PACs to keep U.S. innovation ahead in the AI race.
Diarist Abandons Day One Over Apple AI Journal Privacy Fears ↗
A lifelong diarist ditched the Day One app after Apple revealed its AI-powered Journal at WWDC. The promise of “on-device machine learning” that mines your contacts, photos and habits for writing prompts felt creepy—especially after the Photos app once resurfaced an image of the author’s mother’s open casket. They’d rather face a blank page than an algorithm’s memories.
Palantir’s AI Surveillance Powers ICE Deportations and IDF Strikes, Spurs Privacy Protests ↗
Palantir’s AI-driven surveillance platforms power ICE deportations and IDF strikes, collecting personal data to target and detain migrants and dissidents. Invisible Istar tools—“AI kill chains”—undermine civil liberties, enabling warrantless searches and automated actions without transparency. With US AI consumer protections at risk, activists are protesting Palantir offices from Denver to NYC, warning of a future where data dragnets shape behavior and threaten rights. The author, an ex-Palantir designer, urges renewed privacy laws to curb these intrusive systems before they proliferate unchecked.
OpenAI Warns Investors to Avoid Unauthorized SPVs for Equity Exposure ↗
OpenAI cautions investors to steer clear of special purpose vehicles (SPVs) and other schemes claiming to offer exposure to its equity. In a blog post, the company says unauthorized deals won’t be recognized or hold any economic value, though legitimate offers can exist. As SPVs surge in popularity for pooling funds into hot AI startups, OpenAI and peers like Anthropic are tightening transfer rules to prevent circumvention and protect equity interests.
Hobbyist’s 6GB Victorian AI Accidentally Uncovers 1834 London Protests ↗
A hobbyist developer at Muhlenberg College trained a small AI language model exclusively on 1800–1875 London texts to sound authentically Victorian. In a surprise “factcident,” the model accurately described real 1834 London protests and linked them to Lord Palmerston—events the creator hadn’t explicitly taught it. This tiny AI, built on just 6 GB of data, hints at the power of focused historical LLMs for digital humanities, offering a playful glimpse of linguistic “time travel” through period prose.
Netflix Issues AI Production Guidelines to Safeguard Talent and IP ↗
Netflix has rolled out five “best practices” for partners using generative AI in productions. The rules cover copyright safety, data security, temporary assets, and bans on replacing talent or union work without permission. Low-risk AI uses can proceed after notifying Netflix, but any final deliverables, likenesses, or third-party IP require written approval. Netflix stresses transparency to keep viewers’ trust, even as co-CEO Ted Sarandos hails AI’s potential to enhance creativity and control costs.
Coinbase CEO Brian Armstrong Mandates GitHub Copilot Adoption or Face Dismissal ↗
CEO Brian Armstrong ordered every engineer to onboard GitHub Copilot or Cursor within a week — or face a Saturday review. A few who missed the deadline without a valid excuse were promptly let go. Armstrong’s hard line sends a clear message: AI adoption at Coinbase is mandatory. The company now hosts monthly sessions where teams share creative AI coding tips, though some question long-term codebase maintainability.
Meta partners with Midjourney to license aesthetic AI across its apps ↗
Meta’s new AI chief Alexandr Wang announced a collaboration with Midjourney to license its aesthetic technology for use across Meta’s AI models and products. While terms remain undisclosed, the deal will involve joint research between both companies. Meta plans to integrate Midjourney’s capabilities into its Meta AI app, Facebook posts, WhatsApp and Instagram image generators. Midjourney insists it remains an independent, community-funded research lab, and both firms will reveal more details soon.