Good morning and thank you for joining us again!

Welcome to this daily edition of Beginners in AI, where we explore the latest trends, tools, and news in the world of AI and the tech that surrounds it. Like all editions, this is human curated and edited, and published with the intention of making AI news and technology more accessible to everyone.

THE FRONT PAGE

A New York Bill Could Make It Illegal for ChatGPT to Answer Your Legal Questions

TLDR: A New York bill that just cleared committee would hold AI companies liable for chatbots that give legal, medical, or engineering advice, and critics say it's less about protecting consumers and more about protecting professionals from competition.

The Story:

New York's Senate Internet and Technology Committee unanimously passed Senate Bill S7263 last week. The bill would make AI companies like OpenAI, Anthropic, and xAI legally responsible for any harm caused by chatbots performing tasks that would require a professional license if done by a human. That covers a lot of ground: not just law and medicine, but also dentistry, nursing, pharmacy, architecture, engineering, social work, optometry, podiatry, physical therapy, and veterinary medicine. Under the bill, simply offering legal advice through a chatbot could count as unauthorized practice of law. The most questionable part of the bill? It applies even if companies slap a disclaimer on their chatbot saying "I'm not a real lawyer," it wouldn't protect them from liability.

Its Significance:

Think about what this actually means for regular people. Right now, millions of Americans use ChatGPT or Claude to understand a lease, figure out what a medical symptom might mean, or make sense of a confusing insurance claim. Most can't afford a $300-per-hour lawyer or a specialist visit for every question. This bill wouldn't just limit AI companies. It would cut off one of the most accessible and affordable sources of basic professional guidance that everyday people have ever had. The bill's sponsor, Sen. Kristen Gonzalez, says it's about protecting the public. A policy researcher at the Cato Institute told Reason the bill is likely unconstitutional, comparing it to banning books on law or mental health from libraries. It hasn't passed the full Senate yet, and the companion Assembly bill hasn't left committee. But the fact that it cleared its first vote unanimously in the state with the third-largest economy in the U.S. is a clear indication of where things are heading.

QUICK TAKES

Meta Hit With Class Action Lawsuit Over Smart Glasses Privacy

The story: A class action lawsuit filed in San Francisco federal court accuses Meta of misleading customers about its Ray-Ban smart glasses' privacy features. The suit follows a Swedish investigation that found human contractors in Kenya were reviewing intimate footage from users' glasses, including bathroom visits and sexual encounters, as part of Meta's AI training pipeline. Meta sold over 7 million pairs of the glasses in 2025.

Your takeaway: Meta marketed these glasses as "designed for privacy, controlled by you." The lawsuit argues that's false advertising, since users who activate AI features can't opt out of having their footage sent to overseas workers. The UK's data protection watchdog has already opened an investigation.

OpenAI Launches ChatGPT for Excel and New GPT-5.4 Model

The story: OpenAI released GPT-5.4, its newest model, alongside a beta Excel add-in that lets users build, update, and analyze spreadsheets using plain language. The model uses 47% fewer tokens on some tasks and outperformed office workers 83% of the time on a benchmark covering 44 occupations. New financial data integrations with FactSet, Moody's, and S&P Global are also live.

Your takeaway: Instead of memorizing formulas or debugging errors, you can now describe what you need and ChatGPT builds the spreadsheet for you. It's available now for Plus, Pro, and Business users in the U.S., Canada, and Australia. Google Sheets support is coming soon.

AI Music Startups Suno and Udio Are Trying to Make Peace With the Music Industry

The story: After getting sued by Sony, Universal, and Warner in 2024 for copyright infringement, AI music generators Suno (now valued at $2.45 billion) and Udio are negotiating settlements. Udio has signed licensing deals with Warner, Universal, and indie label Merlin. Suno settled with Warner. Only Sony hasn't reached a deal with either company.

Your takeaway: This mirrors what happened with streaming a decade ago. The music industry fought it, then embraced it on their terms. The question now is whether artists actually benefit from these deals or whether fans will be able to remix and alter songs using AI in ways musicians never intended.

TOOLS ON OUR RADAR

🐧 Audiobookshelf Free and Open Source: A highly capable self hosted podcast and audiobook server that allows you to stream your personal media collection perfectly across all your devices without relying on corporate subscriptions. (Alternative to Audible)

💡 Napkin AI Freemium: An incredibly intuitive visual storytelling platform that instantly transforms your written text into professional diagrams charts and mind maps simply by highlighting the sentences you want to illustrate.

🎬 Fliki Freemium: A highly efficient artificial intelligence studio that converts your blog posts or written scripts into engaging videos complete with incredibly realistic voiceovers and automatically matched stock media.

🧠 Goblin Tools Free: A brilliantly simple collection of task management utilities designed specifically for neurodivergent individuals helping you break down overwhelming chores into tiny manageable steps while estimating exactly how long they will take.

TRENDING

Netflix Acquires Ben Affleck's AI Filmmaking Startup - Netflix bought InterPositive, a stealth AI company Affleck founded in 2022. The 16-person team builds tools focused on visual effects and editing consistency, not generating synthetic performances. Affleck joins Netflix as a senior advisor. The deal came less than a week after Netflix dropped out of the bidding for Warner Bros.

Luma Launches AI Agents That Handle Full Creative Campaigns - AI video startup Luma unveiled "Luma Agents," powered by its new Uni-1 model that works across text, image, video, and audio in a single system. In one demo, the agents turned a brand's $15 million, year-long ad campaign into localized versions for multiple countries in 40 hours for under $20,000. Launch partners include Publicis Groupe and Adidas.

OpenAI Research Finds AI Models Can't Hide Their Thinking - New OpenAI safety research tested whether AI reasoning models can manipulate their own thought process to dodge safety monitors. The answer: not really. Across 13 models tested, controllability scores ranged from just 0.1% to 15.4%. Models actually got worse at hiding their reasoning when given more time to think.

Anthropic's Claude Found 22 Security Bugs in Firefox in Two Weeks - Anthropic's Frontier Red Team used Claude Opus 4.6 to scan Firefox's codebase and discovered 112 bugs, including 14 high-severity vulnerabilities and 22 CVEs total. That's more high-severity bugs than were typically reported in any two-month period in 2025. All fixes shipped in Firefox 148.

AI-Powered T-Shirt Could Detect Hidden Heart Conditions - Researchers at Imperial College London are developing a smart T-shirt with up to 50 built-in sensors that monitors heart rhythms for days at a time. The AI is trained on data from over 1,000 people to detect inherited conditions like Brugada syndrome that standard ECGs often miss. It could be available in hospitals within five years.

Altman Tells OpenAI Staff They Have "No Say" Over Pentagon Decisions - At an all-hands meeting, Sam Altman told employees the Pentagon will listen to OpenAI's technical expertise but doesn't want the company weighing in on whether specific military operations are good or bad. Altman previously admitted the hasty Pentagon deal, announced hours before Iran strikes began, looked "opportunistic and sloppy." He also said he's pushing the DOD to drop its "supply-chain risk" label on rival Anthropic.

TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)

Build a headline generator that applies six proven formulas to any topic and lets you save and rate results

What this does:

Type a topic, pick a number, and hit Generate. The app applies six headline formulas instantly. Save the best ones, rate them with stars, and use Best Practices tips to understand why each formula works.

What this looks like:

WHERE WE STAND(based on today’s news)

✅ AI Can Now: Find more high-severity security bugs in a browser in two weeks than human researchers typically report in two months

❌ Still Can't: Reliably control or hide its own reasoning process, even when explicitly told to try (controllability below 16%)

✅ AI Can Now: Build and update complex Excel spreadsheets from plain language descriptions, outperforming office workers 83% of the time

❌ Still Can't: Replace the creative judgment of filmmakers, which is why Netflix's new AI tools focus on editing consistency and VFX, not writing or acting

FROM THE WEB

Launching Tuesday, March 10th

RECOMMENDED LISTENING/READING/WATCHING

Robo Rally by Richard Garfield (Boardgame)

You program a robot to navigate a hazardous factory floor by playing movement cards in sequence, but everyone else's robots are doing the same thing simultaneously. The result is hilarious chaos as carefully planned routes collide and robots shove each other into pits and lasers.

Thank you for reading. We’re all beginners in something. With that in mind, your questions and feedback are always welcome and I read every single email!

-James

By the way, this is the link if you liked the content and want to share with a friend.

_{Some * designated product links may be affiliate or referral links. As an Amazon Associate, I earn from qualifying purchases. This helps support the newsletter at no extra cost to you and Amazon makes a tiny hair less.}