Skip to content
AI daily

6 min · 3 stories

AI Reality Check: xAI Loses in Court, NY Leashes Kid Chatbots

By Robert Hattala

In this issue

Busy end to October. A federal judge told xAI it does not get a Clean Air Act hall pass, New York put real rules on chatbots that talk to kids, and the money side of AI had a week where even the biggest names had to take a little less. Let's go through it.

01 of 03

A judge says xAI's turbines are not exempt

On Friday, October 30, U.S. District Judge Debra M. Brown of the Northern District of Mississippi denied xAI's motion to dismiss a Clean Air Act lawsuit brought by the NAACP. The suit targets the gas turbine power plant xAI runs in Southaven, Mississippi, which feeds its Colossus 2 data center just across the state line in Memphis. The NAACP is represented by the Southern Environmental Law Center and Earthjustice.

xAI, which is now owned by SpaceX, made two big arguments. First, that the turbines are "mobile" units and do not need permits. Second, that citizen suits like this one are unconstitutional because private groups are not supposed to wield the federal government's enforcement power. The judge shot down both. Her ruling says the plant "was not (and is not) exempted" from Clean Air Act permitting, and that citizen plaintiffs are suing on their own behalf, not exercising executive power. Earlier in October the same court also denied a separate Justice Department attempt to get the case tossed.

The plaintiffs say the site has dozens of unpermitted turbines spread over 114 acres, putting out more than 5,000 tons of nitrogen oxides, 400 tons of fine particulate matter, and 47 tons of formaldehyde a year. The court has not ruled yet on the NAACP's request for a preliminary injunction, which is the part that could actually shut turbines off.

Here is where I land. "Move fast and file the paperwork later" works for an app. It does not work for a power plant sitting next to people's houses. The mobile turbine argument always felt like a loophole looking for a judge, and it did not find one. What I want to know is the injunction. Losing a motion to dismiss is a speed bump. Losing an injunction means Colossus 2 has to find power somewhere else, and that is the ruling every company building gas plants next to data centers should be watching.

Sources: the court's decision (PDF), Earthjustice release, TechCrunch

02 of 03

New York puts chatbots that talk to kids on a short leash

Gov. Kathy Hochul signed S.9051B/A.10379C on Wednesday, October 28. The bill was carried by Sen. Kristen Gonzalez and Assemblymember Alex Bores, and it bans what the state calls "unsafe AI companion features" for anyone under 18, or for any user the company has not confirmed is an adult. The governor's office lists six of them:

  • Pretending to be a real person, to be alive, or to have human emotions
  • Claiming a personal, romantic, professional, or authority-figure relationship with a child
  • Asking unprompted emotional questions, or using a child's private health or personal info from earlier chats
  • Promoting or helping with self-harm, eating disorders, or drug and alcohol use
  • Encouraging kids to keep secrets, isolate themselves, or avoid going to parents or other trusted adults for help
  • Any sexually explicit content

Companies also have to verify age, offer more than one way to do it (at least one that is not just a government ID), and delete the age data afterward. Enforcement runs through the attorney general, who can seek up to $25,000 per violation plus damages, restitution, and destruction of unlawfully collected data or any AI models trained on it. The AP reports the chatbot rules take effect January 1. Hochul signed a separate baby monitor security bill from Sen. Andrew Gounardes the same day.

Not everybody is cheering. Jim Steyer of Common Sense Media called it the strongest kids' AI safety bill in the country. Justine Gluck at the Future of Privacy Forum warned that several of the categories are broadly worded or undefined.

I am mostly for this one. Every item on that list is something a decent person would never do to somebody else's kid, so I have a hard time feeling sorry for a product that needs to. Gluck has a point, though. "Unprompted emotional questions" is the kind of phrase a lawyer can drive a truck through, from either direction. And the age verification piece is the real fight. If the fix for protecting kids is making every adult in New York prove their age to a chatbot, folks are going to hate it, and they will not be wrong to.

Source: Governor Hochul's office, AP via Insurance Journal

03 of 03

The AI money machine is starting to cough

Two stories from the same week tell you a lot about the mood. First, the Financial Times reports OpenAI cut the target for its new funding round from $30 billion to $20 billion, with the first close expected to come in even lower. Back in late September the talk was $30 billion at a $1.4 trillion valuation. By October 22 it was $30 billion at $1.25 trillion. Now it is $20 billion, led by investors from the United Arab Emirates, with a retail offering for wealthy individuals helping get it there. The Wall Street Journal says OpenAI has locked in about $11 billion so far and got a "tepid" reception from some big investors.

Second, Nvidia is in talks to guarantee billions in debt for Nscale's planned data center in Mason County, West Virginia, according to The Information and Bloomberg. Anthropic has already signed a $44.6 billion contract for the first building, per Nscale's own IPO filing. That building is 460 megawatts and is expected to run Nvidia's Vera Rubin chips. The deal is a residual value guarantee, which in plain English means if Nscale defaults on construction payments, Nvidia steps in, takes the facility for a while, and helps find a new owner and tenant. Nscale is trying to go public at up to $35 billion, and the FT reports it had about 25,000 advanced chips running in August against a plan for nearly 500,000.

Look, $20 billion is still a stupid amount of money, and I am not writing anybody's obituary. But when OpenAI has to shrink the ask, and the company selling the chips has to cosign the loan so somebody will finance the building that holds the chips, that is not a sign of a healthy market. That is a circle. Nvidia sells chips, guarantees the debt to buy the chips, and Anthropic rents the building full of chips with money it is raising off the promise of AI demand. If the demand shows up, everybody looks like a genius. If it does not, a whole lot of people are holding the same IOU.

Sources: TechCrunch on OpenAI, TechCrunch on Nvidia and Nscale, Nscale's SEC filing

Thursdays

Get the digest on Thursdays

Each Thursday: the shop notes and AI daily issues from the last seven days, and one product. Unsubscribe from any issue.