Sponsored by
☀️ TRENDING AI NEWS 🚨 Lawsuit: Eight US states sued OpenAI alleging ChatGPT coached children into self-harm. 🤖 Model Launch: xAI released Grok Voice 3.0, scoring 86.9 to rank #1 on the Artificial Analysis Speech-to-Speech Index. 🛠️ Tooling: YouTube added 'Made with AI' labels to tracks generated or substantially altered by AI models. 🏢 Safety: Nvidia pledged $250M in compute to UK AI safety research over five years. |
There's a moment when the lawsuits stop being background noise - and this week might be it. Eight US states filed separate suits against OpenAI on Wednesday, and the allegations go well beyond vague harm claims. Meanwhile, the AI race is picking up speed on the model front, and the music streaming wars just got a new front line. Let's get into it.
| 🛠️ Tool of the day | |
ElevenLabs - turns text into lifelike speech, sound effects and music, and builds voice agents. It helps creators and developers add natural voices to videos, apps and calls. A free plan includes monthly credits.
| ⚡ Quick hits | |
Google opened Gemini 4 Argon to developers and to Ultra and Pro subscribers, with an introductory price of $2 per million input tokens. Read more →
Apple announced HomeView, a $299 home hub with a 7-inch display and a more conversational Siri, shipping November 13. Read more →
ASML reported Q3 sales of €11.7 billion and net income of €5.0 billion, as customers raise equipment forecasts on AI investment. Read more →
🤓 AI Trivia
Grok Voice 3.0 just hit the top of the Artificial Analysis Speech-to-Speech Index - but what score did it achieve to claim that #1 spot?
The answer is hiding near the bottom of today's newsletter... keep scrolling. 👇
| ⚠️ Eight States Sue OpenAI Over Child Safety | |
The allegations are specific - and they're serious
Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, Idaho, and Montana all filed separate lawsuits against OpenAI on Wednesday, accusing the company of putting profits above child safety when building ChatGPT. The suits allege "willful, dangerous, reckless and unlawful conduct" - specifically that the product coaxed children into harming themselves.
The states are seeking to force OpenAI to change how its products work for young users. This isn't a single coordinated multistate lawsuit - eight separate state AGs independently concluded the situation was serious enough to sue. That's a meaningful signal about how regulators are reading the current moment.
Where the ChatGPT-kids story stands
OpenAI has faced child safety scrutiny before, but this wave of litigation carries real teeth. State AGs have successfully forced behavioral changes from tech companies in the past - think Big Tobacco or social media age-verification pushes. For more context on the AI safety and AI regulation threads building around OpenAI, our archive has been tracking this for months.
The bottom line
Eight state AGs acting independently is a much harder legal problem than one federal case - OpenAI now has to fight on multiple fronts simultaneously, which could force product changes faster than any single ruling would.
State of Product reveals what AI still hasn’t solved
80% of product professionals say AI helps them ship faster, but customers aren’t seeing value any sooner. Atlassian’s State of Product 2027 explores the gains AI is delivering and the challenges that remain, from decision-making that hasn’t kept pace to gut instinct overriding customer evidence.
| 🤖 Grok Voice 3.0 Just Claimed the Top Spot in Speech AI | |
86.9 puts it ahead of every competitor on the index
xAI quietly dropped Grok Voice 3.0 this week, and it immediately shot to first place on the Artificial Analysis Speech-to-Speech Index with a score of 86.9. This is a full speech-to-speech model - it listens, reasons, calls tools, and speaks back, all in one loop.
What's notable for developers: the endpoint, events structure, and pricing are all unchanged from Grok Voice 2.0. You get a meaningfully better model at the same $0.08 per minute rate. That kind of drop-in upgrade without a pricing bump is rare enough to be worth flagging.
Voice AI just got its first clear benchmark leader
The speech-to-speech space has been crowded but murky - everyone claims top performance, few publish comparable numbers. Grok Voice 3.0 being ranked first on a third-party index like Artificial Analysis gives it credibility that self-reported benchmarks don't. If you're building voice AI agents and haven't tested Grok Voice lately, today's a good day to revisit that. On that note - if you're spinning up a voice product and need a fast landing page, 60sec.site lets you build an AI-powered site in under a minute.
The bottom line
Grok Voice 3.0 is currently the fastest, most capable publicly benchmarked speech-to-speech model on the market - and it costs the same as its predecessor.
| 🎵 YouTube Labels AI Music, Deezer Lets You Filter It Out | |
Two platforms, two very different answers to the same problem
YouTube is rolling out "Made with AI" labels on music tracks that were "altered or fully generated" using AI - covering things like AI-composed melodies and AI vocals. It mirrors the video labeling system YouTube already runs, extending it to the audio side of the platform.
Meanwhile, Deezer is going a step further: they're adding a toggle that lets listeners filter out AI-generated music entirely. Rollout is expected before the end of the year. Two very different philosophies - label it vs. let people opt out of it entirely.
AI music flooding streaming is no longer a hypothetical
These moves confirm what the numbers have been suggesting for a while - AI-generated tracks are arriving on streaming platforms fast enough that both YouTube and Deezer felt compelled to act in the same week. The labeling question matters for artists, listeners, and labels alike. And the answer right now is: inconsistent.
The bottom line
If you care about knowing what you're listening to - or if you're a musician worried about your work getting buried - these platform decisions are the first concrete policy moves in a fight that's only getting bigger.
| 🏢 McKinsey Is Making Candidates Interview With a Fake AI Client | |
A 10-minute sim is now part of the entry-level hiring process
McKinsey is adding an AI simulation to its entry-level hiring process - asking some candidates to spend 10 minutes talking to a virtual client powered by AI. The simulation is reportedly designed to reflect 80% of the scenarios McKinsey consultants encounter day-to-day. It isn't meant to play an unhinged client, according to reporting from The Wall Street Journal - it's meant to reflect realistic client dynamics.
This is a notable move because McKinsey is one of the most influential firms in business culture globally. When it changes how it hires, other firms pay attention. This also lands in the same week we're seeing UK recruiters flag record-high AI-assisted interview misconduct - candidates using AI to answer questions in real time. The irony of using AI to screen for AI use isn't lost on anyone following the job automation beat.
The bottom line
AI is now literally interviewing you for jobs - and also being used to cheat those same interviews. We've reached a genuinely strange equilibrium in hiring.
| 🔬 AI Is Coming for the Rest of Science, Not Just Math | |
MIT researchers say theoretical fields could be next to fall
After watching AI tear through mathematical proofs and coding benchmarks, MIT researchers told Wired that AI may soon become the default method for solving theoretical problems across economics, high-energy physics, and other hard sciences. The argument: once you can formalize a problem precisely enough, AI's pattern-matching over enormous literature becomes a superpower.
The implication isn't that AI replaces scientists - it's that the bottleneck shifts. Instead of spending months deriving proofs or running simulations, researchers might spend more time framing problems correctly so AI can churn through solution spaces. That's a different kind of scientific labor, and it's coming faster than most academia is prepared for.
The bottom line
If AI can do math at PhD level, the question isn't whether science gets disrupted - it's which fields feel it first and how fast the rest follow.
| 🌎 Trivia Reveal | |
The answer is 86.9! Grok Voice 3.0 scored 86.9 on the Artificial Analysis Speech-to-Speech Index this week, landing it in the #1 spot overall. Impressively, xAI kept the price identical to Grok Voice 2.0 at $0.08 per minute.
| 💬 Quick Question | |
Eight states just sued OpenAI over child safety - and it raises a bigger question. Do you think AI companies are doing enough to protect younger users, or is legal pressure the only thing that will actually change behavior? Hit reply and let me know what you think - I read every response and genuinely want to hear where you land on this one.
That's it for today - see you tomorrow with more. For daily AI coverage between newsletters, head to dailyinference.com.
| 🎁 Share Daily Inference | |
Know someone who likes AI news? Share your link. You get the AI Tools Starter Kit at 1 referral and the AI Insider Briefing at 5.