OpenAI In A Tricky Legal Situation, Ordered To Preserve Private User Data For Court

⚡️ Headlines

🤖 AI

Meta in Talks for Scale AI Investment That Could Top $10 Billion: Meta is exploring a multi-billion dollar investment in Scale AI to strengthen its data annotation and model training capabilities [Bloomberg].

EleutherAI releases massive AI training dataset of licensed and open-domain text: EleutherAI has launched a large-scale text dataset to support transparent and legally compliant AI training [TechCrunch].

Let’s get to the point: Three newsrooms on generating AI summaries for news: News organizations share insights on deploying AI to create efficient, reliable news summaries [Nieman Lab].

Anthropic’s AI-generated blog dies an early death: Anthropic discontinues its AI-authored blog after issues with coherence and consistency [TechCrunch].

ChatGPT Release Notes: OpenAI details the latest updates and feature changes to ChatGPT across its platforms [OpenAI Help].

Apple brings ‘Apple Intelligence’ to the iPhone screen: Apple introduces “Apple Intelligence,” embedding generative AI directly into the iPhone experience [TechCrunch].

Apple lets developers tap into its offline AI models: Developers gain access to Apple’s on-device AI models, emphasizing privacy and performance [TechCrunch].

Google introduces ‘Portrayals’ in Labs to experiment with AI-generated portraiture: Google Labs unveils a new project enabling users to create stylized AI portraits using text prompts [Google Blog].

🦾 Emerging Tech

Apple’s ‘Liquid Glass’ iOS 26 redesign to hint at 20th anniversary iPhone: Apple’s upcoming iOS 26 redesign will feature a liquid-glass aesthetic in anticipation of the iPhone’s 20th anniversary [Bloomberg].

UK appoints first crypto specialist for insolvencies: The UK government assigns its first dedicated crypto expert to handle bankruptcy cases involving digital assets [CoinDesk].

🤳 Social Media

Instagram trialing changes to Reels reach and discoverability: Instagram tests adjustments to boost Reels visibility and user engagement [Social Media Today].

YouTube testing new thumbnail feature to help videos travel globally: YouTube is piloting AI-generated thumbnails optimized for global audiences and cross-cultural appeal [Business Insider].

🔬 Research

Inside the secret meeting where mathematicians struggled to outsmart AI: Mathematicians convened privately to develop methods to challenge and refine AI-generated proofs [Scientific American].

Study quantifies influence of AI suggestions on creative writing: Research shows how AI-assisted suggestions subtly shape creative writing outcomes and authorial choices [PLOS ONE].

The illusion of thinking: A critique of interpretability in current AI models: Apple researchers analyze the cognitive fallacies behind current AI interpretability practices [Apple Machine Learning].

⚖ Legal

Trump’s cybersecurity executive order aims to roll back Biden-era digital programs: A new executive order from Trump targets the dismantling of cybersecurity policies established under Biden [Cybersecurity Dive].

UK ministers delay AI regulation amid plans for more comprehensive bill: The UK government postpones immediate AI legislation, opting instead for broader future reforms [The Guardian].

🎱 Random

Squanch Games reveals ‘High on Life 2’ for winter release: The creators of “High on Life” announce a sequel set for release in winter 2025 [VentureBeat].

🔌 Plug-Into-This

How we’re responding to The New York Times’ data demands in order to protect user privacy

OpenAI has been ordered by a court, at the request of The New York Times, to retain all ChatGPT and API logs—including deleted conversations—indefinitely. While the company must comply for now, it views this mandate as a serious conflict with its privacy norms and is actively appealing the order.

• The NYT and other plaintiffs in a copyright lawsuit seek indefinite preservation of all output log data, even user-deleted chats.

• Affected users include those on Free, Plus, Pro, Team, and standard API plans; only ChatGPT Enterprise, Edu, and Zero-Data-Retention API customers are exempt.

• OpenAI's retention policy—normally 30 days before permanent deletion—is paused until the legal issue is resolved.

• The extra data is stored securely under legal hold, with access limited to a small, audited legal and security team. There is no wider access by NYT or other external entities.

• OpenAI’s leadership, including COO Brad Lightcap and CEO Sam Altman, asserts this directive is excessive and damaging to user trust and precedents around privacy.

🙈 OpenAI is being forced into a tough legal position, and there’s no doubt the New York Times would like nothing more than to see everyone’s data aired out in the open. Kind of a dirty tactic, but it’s activating one of the key concerns underlying OpenAI (and Big Tech as a whole)…privacy.

Apple’s AI‑driven Stem Splitter audio separation tech has hugely improved in a year

Over the past year, Apple has significantly upgraded its AI-based “Stem Splitter” feature in Logic Pro, dramatically improving its ability to separate audio tracks into vocal, instrumental, and ambient layers. These enhancements reflect both signal-processing refinements and neural model advances.

• Apple’s stem separation, which deconstructs mixed audio into individual components, now offers much clearer isolation of vocals, drums, bass, and other tracks.

• Compared to the previous year, it delivers noticeably cleaner, more distinct stems with reduced artifacting.

• While proprietary, its underlying approach aligns with broader research trends in Music Source Separation using neural networks—often trained on large multitrack datasets.

• The refined Stem Splitter empowers musicians and producers to remix, restore, or learn from existing recordings without need for original multitracks.

• With improvements in separation fidelity, it now supports more professional-level use cases like sample extraction and forensic audio cleanup.

🎼 Simply put: Apple’s Stem Splitter is getting significantly better at unmixing songs—like pulling out each instrument or vocal track clearly from a blended recording. With this leap, Apple is reinforcing Logic Pro as a serious tool for audio professionals and hobbyists alike, blurring the lines between AI-aided editing and creative control.

Like humans, AI is forcing institutions to rethink their purpose

As AI progresses, institutions—from universities and governments to corporations—must reexamine their original purpose. The shift is akin to a “cognitive migration,” requiring structural reinvention and deeper alignment with values that machines cannot quantify.

• The concept of "cognitive migration" posits that institutions, like individuals, must adapt their mental frameworks and roles as AI reshapes reasoning, judgment, and coordination.

• AI’s rapid integration compounds existing stressors on institutions: declining trust, political fragmentation, and outdated processes.

• Many legacy institutions were architected for human-centric cognition—stable hierarchies, expert-driven decision-making, and centralized authority—and these models are under pressure.

• Some are preemptively experimenting: educators are shifting from content delivery to mentorship, reframing their roles as AI reshapes teaching.

• Ultimately, institutions must grapple with whether to resist, retrench, or reinvent in response to AI's capabilities, while preserving human dignity, ethics, and long-term social purpose.

🔍 This invites a deeper conversation about institutional resilience—not just rapid tech adoption, but values-oriented reinvention. The biggest question: can slow-moving structures truly evolve in time when the rate of change is accelerating?

 🆕 Updates

📽️ Daily Demo

🗣️ Discourse