OpenAI’s big privacy problem

Good morning, AI enthusiasts. OpenAI’s NYT copyright battle just became a privacy nightmare — with a court order forcing the AI leader to retain tens of millions of personal (and even deleted) user conversations.

With CEO Sam Altman proposing “AI privilege” and the tech moving towards deeper digital relationships with always-listening devices, this legal fight may set the stage for a brand new era of digital rights.

In today’s AI rundown:

• OpenAI fights court to preserve convos

• AI policy head weighs in on human-AI bonds

• How you can turn complex research into quizzes

• AI reveals Dead Sea Scrolls are a century older

• 4 recent AI tools & 4 job opportunities

LATEST DEVELOPMENTS

OPENAI

🔒 OpenAI fights court to preserve convos

The Rundown: OpenAI is pushing back against a recent court order stemming from its ongoing legal battle with The Latest York Times, which forces the corporate to retain all user conversations — including deleted chats.

The main points:

• The mandate affects lots of of tens of millions of ChatGPT users across its free, Plus, Pro, and Team tiers, forcing OpenAI to retain even manually deleted chats.

• The Latest York Times argued for the information preservation out of concern that users may be infringing on its content after which deleting the evidence of their chats.

• CEO Sam Altman called the demand an “inappropriate request that sets a nasty precedent” and proposed “AI privilege” just like doctor-patient confidentiality.

• ChatGPT Enterprise, Edu, and API customers that use a Zero Data Retention agreement are excluded from the court order.

Why it matters: This order arrives at a moment when tens of millions are beginning to trust AI with vulnerable thoughts and knowledge (for higher or worse) — threatening to shatter trust just because it becomes embedded in day by day life. The privacy query will develop into even thornier as AI transitions to physical devices with always-on recording capabilities.

TOGETHER WITH CONVEYOR

🛠️ AI was made for this

The Rundown: It’s 2025 — security questionnaires should complete themselves, your customers should access your Trust Center immediately, and your content library should not be a maintenance burden Sue and Phil, Conveyor’s AI Agents, deliver exactly that with 95%+ accuracy.

Let Sue and Phil do the busy work so you possibly can:

• Spend 90% less time answering security questionnaires and RFPs

• Skip the manual updates—your content library stays current by itself

• Give your customers easy AI answers in your Trust Center

• Fully automate admin and collaboration tasks across teams and tools

Learn more and put Conveyor’s AI agents to work.

OPENAI

🤗 AI policy head weighs in on human-AI bonds

The Rundown: OpenAI Head of Model & Behavior Policy Joanne Jang just published a blog post detailing how the corporate is approaching human-AI relationships and questions of AI consciousness.

The main points:

• Jang said that individuals naturally anthropomorphize AI, an inclination amplified by models responding with what appears like non-judgmental empathy and validation.

• OpenAI considers AI consciousness currently unanswerable, as an alternative specializing in how conscious it appears to users and its impact on mental wellbeing.

• The design philosophy is to string a nice needle, aiming for a personality that’s warm and helpful without giving it a fictional backstory, feelings, or desires.

• Jang also said that evolving human-AI relationships reflect each how people use the tech, but in addition “may shape how people relate to one another.”

Why it matters: AI labs have gotten the architects of digital companionship for tens of millions of users, making decisions that can shape societal norms around connection and empathy. This uncharted territory is actually one big social experiment — and even the creators are uncertain concerning the psychological territory we’re entering.

AI TRAINING

📚 How you can turn complex research into quizzes

The Rundown: Google Gemini’s deep research feature can now analyze any topic in depth and mechanically turn findings into interactive quizzes perfect for studying or teaching.

Step-by-step:

1. Visit Google’s Gemini website and click on “Deep Research” at the underside of the chat interface

2. Enter your educational topic with specific requirements (grade level, focus areas), and review and edit the research plan as needed

3. Let Gemini generate a comprehensive report from multiple sources (takes a couple of minutes)

4. Click “Create Quiz” to mechanically generate an interactive quiz with multiple query types

Pro tip: The quiz will offer you explanations after you answer either correct or incorrectly, review them after your mistakes to learn faster.

PRESENTED BY RUNNER H

🚨 Your job just got 1000x easier

The Rundown: Runner H is the latest AI agent you possibly can delegate all of your boring, repetitive, and time-consuming tasks to! Give it access to your tools, and it will probably handle entire workflows from a single prompt.

Some tasks you possibly can delegate to Runner H include:

• Reading your essential emails and drafting (and even sending!) replies

• Making a Google Sheet with trending ad ideas and sharing it along with your team on Slack

• Finding job opportunities and applying for you

Watch this video to see Runner H in motion, before trying it at no cost to experience an agent that transforms the way in which you’re employed!

AI RESEARCH

📜 AI reveals Dead Sea Scrolls are a century older

The Rundown: Researchers just discovered that the Dead Sea Scrolls could also be as much as a full century older than previously estimated after training an AI system called Enoch to research ancient patterns and mixing it with radiocarbon dating techniques.

The main points:

• Enoch was trained by linking known radiocarbon dates of scroll fragments with handwriting styles, learning to associate visual patterns with time periods.

• The brand new dating pushes some biblical texts back to the time of their presumed authors, with some texts coming in at as much as 2,300 years old.

• The AI method offers a non-destructive alternative to carbon dating, which requires cutting samples from the dear manuscripts.

Why it matters: Even fields like archeology are seeing transformations within the AI era — with pattern recognition abilities continuing to unlock recent secrets from history (like deciphering the Herculaneum scrolls). With ancient texts stored across museums and libraries worldwide, there remains to be tons of data waiting to be discovered.

QUICK HITS

🛠️ Trending AI Tools

• 🤖 Gemini 2.5 Pro – Latest update with improvements across benchmarks

• 🗣️ Eleven v3 – SOTA text-to-speech model with support for 70+ languages

• 👄 Bland TTS – Latest voice AI with enhanced realism and control

• 🎥 HunyuanVideo-Avatar – Create multi-character talking videos from audio

💼 AI Job Opportunities

• 🤝 The Rundown – Partnerships Manager

• 🎤 Parloa – Lead Product Trainer

• 📅 UiPath – Executive Assistant

• 🧱 OpenAI – Full Stack Engineer, Post Training

📰 All the pieces else in AI today

Apple researchers published a brand new study revealing that reasoning models hit a “scaling limitation” where they think less and perform worse as complexity increases.

Anthropic added national security heavyweight Richard Fontaine to its Long-Term Profit Trust, deepening the corporate’s give attention to navigating AI’s global risks.

OpenAI rolled out an update to its Advanced Voice Mode, featuring more natural, expressive speech and improved translation capabilities.

Anysphere released Cursor v1.0, with recent features including a Background Agent for distant coding, BugBot for automatic PR review, and recent memory capabilities.

Google launched Portraits, a Labs experiment allowing users to have personalized experiences with AI versions of experts based on their voice and knowledge base.

Higsfield AI introduced Higgsfield Speak, a brand new update enabling talking avatars with custom styles, scripts, and motion.

FutureHouse released ether0, an open-weights chemistry-focused reasoning model that significantly outperforms top models on scientific tasks.

COMMUNITY

🎥 Watch our last live workshop

Try our last workshop with Dr. Alvaro Cintas, The Rundown’s AI professor. By the tip of the workshop, you’ll confidently have the opportunity to include Flux Kontext into your creative workflows for consistent, high-quality results.

Watch here. Not a member? Join The Rundown University on a 14-day free trial.

See you soon,

Rowan, Joey, Zach, Alvaro, and Jason—The Rundown’s editorial team