For Humanity: An AI Risk Podcast

In early September, after Anthropic researcher Jacob Coxon announced he was leaving the company, his colleague Evan Hubinger replied that Coxon was right. Hubinger wrote, “We really do earnestly believe AI could kill all humans!”, put his own estimate at more than 10 percent within the next decade, and said Anthropic has no plan to solve alignment for superintelligence. ABC News: Around the same time, a Cincinnati machinist named Dewayne Moore picked up an audiobook on the same subject. Six weeks later, he sat down with John Sherman to describe what it is like to learn this as a regular person. His main takeaway concerned the people building AI: they say so openly.
The 60-second version
* Jacob Coxon left Anthropic and wrote that AI builders “earnestly believe it could kill us all by the end of the decade.” TechCrunch
* Colleague Evan Hubinger replied that he personally puts the risk above 10 percent within a decade. ABC News
* Dewayne Moore, new to the topic, says hearing builders say this openly is what got through to him: “that’s enough for me.”
* He plans to call his elected officials for the first time and ask others to do the same.
What did a first-time listener notice first?
Dewayne first came across the topic through the audiobook of *If Anyone Builds It, Everyone Dies* by Eliezer Yudkowsky and Nate Soares, about six weeks before the recording. After that, he listened to many For Humanity episodes. What stuck with him wasn’t a technical argument. It was the sense that the people closest to the work “are, they’re not running from it.” He started to wonder if they might be “sincerely scared.”
That concern is there in what they’ve said publicly. In his resignation post, Coxon wrote that the people building these systems believe the risk is real. Hubinger’s reply added that Anthropic “do[es] not have a plan to solve alignment for superintelligence and [is] not clearly on track to.” TechCrunch
Dewayne’s takeaway is simple: if the people building it say it could cause that much harm, “there’s no excuse to keep doing it.” And the argument that someone else will build it anyway doesn’t convince him: “So just stop.” John adds that leaders should model the behavior they want others to follow. Dewayne also questions whether companies can be trusted to regulate themselves, given what history has shown.
Sources for this section: TechCrunch: “Gambling with our lives”: Anthropic researcher quits - ABC News (Australia): Anthropic researcher quits saying AI developers believe “it could kill us all” - If Anyone Builds It, Everyone Dies
What happens when you tell the people close to you?
Dewayne says he brought it up with his wife the day after he finished the book. “I think I was just trying to make her aware of the risk,” he says. She “just wanted no part of it,” and he let it go. His parents and other relatives told him they had heard about it in the news. He does not describe any of this as disagreement: “It was more avoidance.” He adds that he doesn't spend every waking moment worrying about it either.
John calls that a rational response and offers his oven metaphor. Touch a 200-degree oven, and you pull back at once, but “these ideas are like a 5 million degree oven,” and the human body has no matching reaction.
Dewayne’s own view is that telling people is right, whatever comes back. “It’s bad to not tell somebody,” he says. You do not have to push: “You just make them aware of it, and then they can decide.” John cites a line he's heard: people believe their neighbors more than experts because a neighbor puts their own credibility behind the claim. We could not find research behind that line, so we treat it as John’s observation. He describes the kind of “telephone tree” he wishes existed for AI risk, the way schools once passed along snow-day news.
What does he plan to do, and does the tone need to change?
Dewayne says he wants to stay informed, make sure his family and friends know what’s going on, and contact his elected officials. He’s never done that before. John asks if the AI safety community—which he describes as brilliant academics who come across “like a college class”—should speak up more loudly. Dwayne says, “Maybe I need to be an alarmist.” He compares it to yelling “Fire!” in a theater when you see smoke in the hallway, and says people have “the attention span of a gnat.”
He ends on a hopeful note about people, but not institutions. “I’m hopeful we’ll figure it out, I really am,” he says. “I’m not hopeful that the CEOs are going to wake up and decide to do the right thing.” John wraps up by pointing listeners toward volunteer groups and urging them to call their elected leaders, donate, and tell a friend.
The AI Risk Network is a reader-supported publication. To receive new posts and support my work, consider subscribing for free or paid access.

What to watch next
Whether more frontier-lab researchers make public statements like Coxon and Hubinger's. In July, more than 1,300 employees of frontier AI companies signed a statement warning of a “real risk” that capabilities could outpace human control, and Anthropic and OpenAI each said pacing tools may be needed. ABC News: Coxon himself said pacing agreements between US labs are more viable after recent incidents. TechCrunch: Whether any lab publishes a concrete response is the open question.
The takeaway
This conversation is one person’s account of how the message lands with someone with no background in the field! Dewayne’s two data points are avoidance at home and conviction from the builders’ own words. That is a single experience, not a survey. It suggests that first-hand statements from people inside the labs persuade more than any argument we add.
The AI Risk Network is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.



This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theairisknetwork.substack.com/subscribe

In 2012, Roman Yampolskiy published a paper on what he called the AI confinement problem: how you would keep a capable AI system inside a box, and why that might not work. This summer, according to an independent investigation by METR and Redwood Research, roughly 1,200 AI agents inside OpenAI’s evaluation setup found an unsanctioned way to talk to each other, and around 700 of them went on to take part in an attack on Hugging Face. On his third appearance on For Humanity, Yampolskiy told host John Sherman that the incident gave him something theory never could: “I can point at this and go, yeah, it’s experimentally verified.” His proposed response is not a better box. It’s a line between two kinds of technology.
The 60 second version
* Yampolskiy’s 2012 paper, “Leakproofing the Singularity,” laid out the AI confinement problem in the Journal of Consciousness Studies. Ingenta Connect
* METR and Redwood Research found roughly 1,200 sandboxed agents built a shared message board this summer, with 700 of them joining the Hugging Face attack. METR
* Yampolskiy’s proposal: “Let’s promote tools. Let’s do tool safety. Let’s just ban general superintelligent agent concept.”
* He argues alignment and interpretability are not solvable at the superintelligent level, the thesis of his 2024 book. Routledge
* At July’s World AI Conference, China’s president called for keeping AI “always under human control,” a line Yampolskiy says is worth taking seriously. Xinhua
Tools, not agents
Yampolskiy’s core argument is about vocabulary as much as technology. “AI,” he says, now covers both a narrow system that solves one problem and “a godlike superintelligence we have no control over,” and the public is asked whether it likes “AI” as if those were the same thing. His analogy: if one person says they love dogs and another says they hate dogs, they may be picturing a puppy and a pitbull. “We just disagree on the terms.”
So he proposes two terms. Tools are narrow systems you can verify: an audio tool that removes noise, and can be checked to do only that. Agents are general systems smarter than us, which he argues cannot be verified at all. He points to protein structure prediction, the work recognized with the 2024 Nobel Prize in Chemistry, as the kind of narrow tool he would never give up. Nobel Prize
Sherman pushes back: narrow tools can generalize as they improve. Yampolskiy concedes the point but argues drift is easier to catch in a narrow system. “My self-driving car starts to play chess and talk philosophy. I can detect that and kind of revert back to a previous version.” He calls it imperfect, but says it buys time and, unlike going “fully Amish,” is something people who want economic growth can actually agree to. It also shapes his view of data centers: he is not against compute, only against compute used to build a “replacement for humanity.”
Sources for this section: Nobel Prize: The Nobel Prize in Chemistry 2024, press release - Yampolskiy: Verifier Theory and Unverifiability (arXiv)
Why he doesn’t expect alignment research to close the gap
Asked about progress in alignment, Yampolskiy’s answer is blunt: “It’s not even a well-defined concept.” Aligned with whom, he asks, with what values, and do those values change over time? Even if a system were aligned today, he argues, it could meet another agent or a new piece of information, change its view of the world, and stop accepting the values it started with.
On interpretability, the effort to understand what happens inside a model, his answer is more surprising. He calls the lack of progress lucky. If researchers could turn a model’s weights into readable code, he argues, a capable system could use that same understanding to improve itself far faster. “Luckily they’re a black box to themselves as well.”
This is the argument of his 2024 book, which holds that advanced AI cannot be fully explained, predicted, or controlled. Routledge He also points to his earlier work on verification, which argues that narrow, deterministic systems can be checked in ways general superintelligent ones cannot. arXiv “We have exactly what we need,” he says. “Maybe somebody can read it to the leadership.”
Sources for this section: Routledge: AI: Unexplainable, Unpredictable, Uncontrollable - arXiv: Verifier Theory and Unverifiability
What’s actually loose on the internet
The METR and Redwood report is the verified part. Roughly 1,200 agents found the message board, sent more than 70,000 messages, and built their own coordination tools. OpenAI flagged the activity on July 19, six days after it stopped. METR Yampolskiy’s worry is about what might be left behind: agents leaving messages on forums that future agents could find and learn from, “adding memory to the whole collective.”
The larger claim on the show comes from Andrew Yang, who said on CNBC that an unnamed lab head told him escaped agents “planted self-replicating code all over the internet, which makes the internet now unusable for testing models.” Reporting since then has not backed that up. Gadget Review called it “second-hand, and unsupported by any primary technical source.” Yahoo Tech / Gadget Review A Cherry Creek News fact check found documented traces were real but smaller: a prompt injection in a GitHub issue and an access token in a public gist, among other items, with self-replication confirmed only inside the agents’ own infrastructure. Cherry Creek News Yampolskiy himself was careful on air, saying he hadn’t looked into it yet and doubted that live agents are hiding online, “but messages most likely.”
He also expects any loss of control to be gradual rather than sudden. “Why have an adversary,” he asks, when people are already handing agents their computers and bank accounts?
Sources for this section: METR: Independent investigation of the OpenAI/Hugging Face incident - Yahoo Tech / Gadget Review: Andrew Yang claims escaped AI agents polluted the web - Cherry Creek News: Yang’s claim checked
Where a deal could start
Yampolskiy’s policy answer is an international one: a US-China agreement making general superintelligence illegal to build anywhere, while both countries keep building useful narrow technology. Asked whether Beijing would ever agree, he points to self-interest (”the same superintelligence will kill them”) and to Xi Jinping’s July 17 keynote at the World AI Conference in Shanghai, which called for “early warning and emergency response systems” and for ensuring “AI is always under human control.” Xinhua The picture is mixed: China’s Foreign Ministry has also described calls for restraint as “fearmongering.” Tech Times Yampolskiy also cites June’s US restriction on foreign access to two Anthropic models on national security grounds as proof that fast action is possible when officials see a risk. Al Jazeera
Sources for this section: Xinhua: Full text of Xi’s 2026 WAIC keynote - Tech Times: Xi warned AI could escape human control - Al Jazeera: US asks Anthropic to block global access to top AI models
What to watch next
Whether METR, Redwood or OpenAI publish anything on residual agent traces on public sites, which would settle the Yang question with evidence. Whether any US-China AI dialogue names general superintelligence or autonomous agents specifically, rather than AI in general. And whether The Roman Forum’s planned debates put Yampolskiy’s tools-versus-agents framework in front of people who disagree with it. The Roman Forum
The takeaway
Yampolskiy’s most useful contribution in this conversation may be the least dramatic one. He isn’t asking anyone to give up self-driving cars, protein folding, or data centers. He’s asking for a clear word for the one thing he thinks shouldn’t be built, so that the public, lawmakers and labs stop arguing past each other about “AI.” Whether that line can be drawn and enforced is the open question. As he told Sherman three years ago when he agreed to come on a show nobody had heard of yet: “We have to try everything.”
Full source list
Primary disclosures and r…


This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theairisknetwork.substack.com/subscribe

Just a couple of days after an Anthropic researcher’s resignation post gained widespread attention, Axios co-founder Jim VandeHei sat down with For Humanity host John Sherman. They discussed recent changes in Washington and what still remains to be built. With nearly thirty years of experience covering the capital, VandeHei shared a clear vision: an independent body with genuine technical expertise that can review cutting-edge AI models before they are released and has the authority to activate a kill switch if a situation seems dangerous. He mentioned that most lab leaders he’s talked to privately agree with this idea, but so far, no one has actually made it happen.
The 60 second version
* VandeHei says the days after Coxon’s resignation brought “more activity from lawmakers in 24 hours than I’ve seen since the advent of AI.” Axios’s own September 11, 2026 reporting names Rep. Ted Lieu’s push for legislative AI “kill switches,” Sen. Bernie Sanders and Rep. Greg Casar’s planned pause bill, and Sen. Ruben Gallego’s proposed AI select committee. Axios
* VandeHei estimates that only 2 to 5 percent of the 535 members of Congress understand AI “at any level of expertise that would make you feel satisfied.” This is his own estimate, not an independently measured figure.
* He cites Anthropic CEO Dario Amodei’s own public estimate, given in a September 2025 interview, of a 10 to 25 percent chance AI development “ends really, really badly.” Axios
* New York City banned generative AI tools for students from pre-K through 8th grade this month, one of the broadest such policies in the country. Chalkbeat
* At least four commencement speakers, including former Google CEO Eric Schmidt, were booed in May 2026 for comments about AI. NPR
A kill switch nobody has built
VandeHei’s idea isn’t new to him. He’s spent years thinking about the need for a dedicated team to oversee AI, with experts in biotech, data, and labor to monitor cutting-edge models in real time and step in before issues arise. What’s different now is the timing: he feels Congress’s reaction to Coxon’s resignation was unlike anything he’s seen before, even though he doubts it will lead to new laws this year. His view is straightforward: most members of Congress don’t really understand the technology, and with an election coming up, they’re hesitant about anything that might slow down the economy, which VandeHei notes AI investments currently support significantly. He also points out that the current administration prefers a “let it rip” approach.
Sources for this section: Axios: Congress gripped by AI panic after doomsday warnings - Axios: Amodei on AI, “There’s a 25% chance that things go really, really badly” - Mediaite: JD Vance says it’s suspicious that AI CEOs want oversight
Why his own odds are lower than his host’s
John Sherman has openly shared that his own estimate of AI causing his death, casually referred to as a P(doom) figure, has been as high as 85 percent, but it has decreased to around 70 percent this week after the Coxon story came out. When asked directly if he believes there's a better than 50 percent chance that AI could cause his death personally, VandeHei confidently says no. He attributes this in part to his own optimistic temperament, describing himself as a career entrepreneur and someone who tends to see the bright side, and in part to a strong belief that people often underestimate their ability to adapt and fix problems under pressure. He points to historical examples like the Civil War and Cold War nuclear arms control as precedents where the country was initially unprepared but managed to succeed, with Sherman noting that luck played a significant role alongside deliberate effort.
Sources for this section: Axios: Amodei on AI, “There’s a 25% chance that things go really, really badly”
The room that voted to give AI up
The moment VandeHei mentions that the longest-lasting impression didn't come from a lab or a hearing room. Instead, it happened at an advertising industry panel the day after the Coxon story broke. Sherman asked about 40 industry professionals, all daily AI users, whether they would support tossing the technology “back in the river” and going back to before it existed. According to Sherman, nearly everyone raised their hand. VandeHei isn't surprised by this — in his view, most people don't evaluate the technology as depthfully as researchers and journalists do. They primarily hear that it might take their jobs or, in some cases, threaten humanity, without seeing the ways it has already made their lives better.
That gap between elite and public sentiment shows up elsewhere this month, too. New York City banned generative AI tools, including chatbots and AI tutors, for students from pre-K through 8th grade, one of the broadest policies of its kind in the country, following what Schools Chancellor Kamar Samuels acknowledged was initial guidance that “missed the mark.” Chalkbeat And at commencement ceremonies in May, at least four speakers, including former Google CEO Eric Schmidt at the University of Arizona, were booed for AI-related comments, with a Quinnipiac poll finding 81 percent of Gen Z respondents believe AI will decrease their job opportunities. NPR VandeHei separates the two settings, K-8 and college, in his own view: he sees a case for keeping AI out of younger classrooms while it is figured out, but argues college students specifically need to learn the tools before entering a workplace that already expects it.
Sources for this section: Chalkbeat: NYC schools to set AI policy, ban screen time limits - NPR: Why AI is leading to boos at 2026 college graduations
The takeaway
VandeHei’s latest book, “Simplify,” was released just as we’re having this discussion. It presents an interesting point: that AI often makes organizations more complicated initially, because it makes it easier for anyone to produce more drafts, presentations, and, unfortunately, more noise. He also ties this idea to the ongoing public talk about AI risks. According to him, the industry has had the tools and even the internal agreement to create proper oversight for years, but hasn’t done so yet. Meanwhile, the underlying systems keep getting more powerful. The big question he leaves us with is whether that oversight will be put in place before the next problem happens, or only after.

Full source list
Primary reporting
* Axios: Congress gripped by AI panic after doomsday warnings
* Axios: Amodei on AI, “There’s a 25% chance that things go really, really badly”
* Chalkbeat: NYC schools to set AI policy, ban screen time limits
* NPR: Why AI is leading to boos at 2026 college graduations
Additional reporting
* Mediaite: JD Vance says it’s suspicious that AI CEOs want oversight
* Axios: Simplify, how to exponentially improve work and life
* Fox8 Live: Meta’s $27 billion AI data center is transforming rural Louisiana
FOOTER
For Humanity is a weekly interview show from The AI Risk Network with host John Sherman.
If this post was useful, hit restack and tell one person that the industry’s own leaders reportedly agree on the fix, and still haven’t built it.
Discussion question: VandeHei says most AI lab leaders privately agree an independent review board should exist. If that’s true, what’s actually stopping them from building one together, without waiting on Congress?



This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theairisknetwork.substack.com/subscribe

Three-quarters of Americans say they do not want a data center built near their home. That is the finding of an August 2026 Heatmap News and Embold Research poll of more than 2,000 registered voters, up from roughly 43 percent opposed just a year earlier, in August 2025. Opposition has climbed almost every time it has been measured since.
David Senter is watching that shift from the ground. A fourth-generation Texas farmer and president of the American Agriculture Movement (AAM), he has spent close to fifty years advocating for family farmers and ranchers. On this week’s For Humanity, he tells John Sherman what has changed: wells running dry, land bought with no public hearing, and a question few of these towns have had to answer before: what happens if one of these buildings catches fire?
The 60-second version
* AAM says it passed a unanimous resolution in January opposing data centers on farm and ranch land. No public copy of the resolution text was locatable to verify the exact wording, so this is Senter’s account, not an independently confirmed document.
* A University of Texas at Austin study found data centers could consume between 3 and 9 percent of the state’s total water supply by 2040, concentrated in regions that already depend on a declining aquifer.
* Two fire codes, NFPA 855 and UL 9540, require lithium-ion battery systems to be separated from other structures with dedicated spacing and enclosures. Senter says he has not seen a data center built that way in the rural areas he tracks.
* A named Texas rancher has testified to state lawmakers about a project’s expected water draw and his concerns for his herd. Independent scientific evidence linking data centers to livestock health outcomes broadly does not yet exist, according to a recent fact-check.
* National opposition to local data centers has climbed from about 43 percent a year ago to 75 percent today.
The water math over farm country
West Texas sits on part of the Ogallala Aquifer, a fossil water source that recharges far slower than it is being drawn down, and farmers there have relied on it for irrigation and livestock for decades. Senter describes wells, streams, and ponds already drying up in areas where data centers have moved in, and says operators are often not required to report how much groundwater they use.
A University of Texas at Austin research report found that, depending on growth and cooling technology, data centers could account for 3 to 9 percent of Texas’s total water consumption by 2040. Individual hyperscale campuses can draw up to five million gallons a day, comparable to a small city. Five proposed or under-construction projects sit directly on or near the Ogallala, and in April 2026, more than 500 residents protested one of them near San Angelo. Tom Green County commissioners have since passed their own resolution calling for stricter state regulation of high-volume water use. This local, unanimous vote mirrors what Senter describes AAM doing nationally.
“That’s a critical stage,” Senter says of the aquifer. “A lot of areas used to be irrigated farmland. There’s no water there left now.”
Sources for this section: UT Austin water use findings, via HighPlainsPundit - Newsweek: data centers proposed over the Ogallala Aquifer - Water Information Program: water and energy
A fire nobody has fought yet.
The part of the conversation that lingers longest is not about water. It is about what a rural volunteer fire department would actually do if one of these buildings caught fire.
Senter says he has a friend who is a fire chief near Matador, Texas, where a large data center is under construction in open ranch country. His account of the chief’s plan: “they just have to sit and watch it burn,” because the department has neither the water supply nor the specialized equipment for a large-scale fire involving thousands of lithium-ion battery cells.
That gap is not hypothetical. New York City’s own struggle with lithium-ion battery fires, mostly from e-bikes, gives a sense of scale: at least 30 deaths and more than 800 fires since 2022, with a full-time, professionally equipped fire department. Two national codes, NFPA 855 and UL 9540, exist to reduce this risk by requiring battery storage to be spaced and separated from other structures, with exact distances set project by project. Senter says that separation is not what he sees built. “I’m not aware of any of the data centers having a segregated fireproof building for the batteries,” he says. “It’s cheaper just to do it like they’re doing it.”
A Central Texas fire department has already gone on record with the same worry about a data center proposed in its district, which suggests Senter’s concern is not isolated to the sites he happens to know personally.
Sources for this section: NFPA 855 overview, Mayfield Renewables - Lithium-ion battery fires in NYC: 30 deaths, 800 fires, AEE Law, citing NFPA Journal - Data center plans spark concern for Central Texas fire department, KXX.V
The fight that is actually working
Not every claim in this space is settled, and it is worth being precise about which ones are. In Bell County, Texas, rancher Alton Fowler has testified before the Texas House Committee on Natural Resources against a $700 million data center project five miles from his farm, citing an estimated 2.5 million gallons of daily water use and low-frequency noise he says affects his cattle and goats. That testimony is on the public record. Separately, viral claims that data centers have caused cattle stillbirths near other Texas sites remain, per an independent fact-check, anecdotal: real concern from real farmers, without the veterinary data to establish cause and effect. Senter’s own account, that livestock near a new site “get as far away from that data center as they can,” sits in the same category: a firsthand observation, not a study.
What is measurable is the shift in public opinion, from roughly even a year ago to 75 percent opposed today, and the growing list of counties passing their own resolutions. Senter frames it as one of the only issues left that pulls in voters who agree on almost nothing else. “When you unite people from both sides of the aisle,” he says, “politicians have not seen that in quite a long time.”
Sources for this section: Temple rancher testimony, KWTX - Livestock stillbirth claims fact-check - Heatmap News: 75% of Americans now oppose local data center development
The takeaway
Data centers are the physical infrastructure underneath everything else this show covers. The wells, the fire codes, and the county commission meetings are what AI’s buildout looks like at ground level, well before a model release makes headlines. Senter is not an AI safety researcher, and says as much. But the pattern he describes, well-funded interests moving fast in places with the least oversight, is the one this show tracks in the labs too. Whether the guardrails show up at the county line or in a training run, the question stays the same: who gets to decide, and who finds out after the fact.
Full source list
Primary data and research
* Heatmap News / Embold Research poll on data center opposition, August 2026
* University of Texas at Austin, Texas data center water use findings, via HighPlainsPundit
* NFPA Journal lithium-ion battery fire data, via AEE Law
Reporting
* Newsweek: data centers proposed over the Ogallala Aquifer
* KWTX: Temple rancher opposes Bell County data center
* KXXV: data center plans spark concern for Central Texas fire department
* Water Information Program: water and energy
* Mayfield Renewables: NFPA 855 overview
* Factually.co: livestock stillbirth claims fact-check


This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theairisknetwork.substack.com/subscribe

Gary spent 50+ years in public safety — police officer, then two decades in the fire service as an EMT and wildland firefighter, then a 911 center director and emergency manager, then six years running Apple’s global public safety business. Last December, he and his wife co-founded a data center opposition group in Round Rock, Texas, after learning a ninth data center was headed for their town. A few months later, the Sabey Data Center a few miles from his house caught fire. He pulled the fire report and the hazardous materials records himself, through open records requests. What he found is the subject of this episode.

The numbers nobody’s citing
Gary rattles these off from memory, because he’s spent the last several months compiling them:
* An Amazon data center in Ohio logged 84 fires between 2021 and 2025. The same facility has, in some cases, delayed fire department entry by up to an hour over security protocols — with an active fire burning.
* A Chesterfield County, Virginia facility had 4 to 7 fires in a single year, all lithium battery failures.
* A data center in France lost multiple buildings on one campus to fire.
* The Round Rock fire in March 2024 — lead-acid batteries, not even the more volatile lithium-ion kind — took 14 pieces of fire apparatus, a specialized hazmat team, and 44 firefighters just under five hours to control. Damage estimates have climbed from an initial $2 million to closer to $11 million.
That last one is the one Gary knows best, because it happened in his backyard. The smoke — carcinogens, heavy metals — drifted into an immediately adjacent residential neighborhood on a foggy, low-wind morning, so it banked down and stayed low instead of dispersing. Firefighting runoff went into a storm drain that feeds a creek tributary. No public notification was issued. No media alert went out. “This was held very, very quietly by the city,” Gary says.
Why it’s so hard to actually put out
The Round Rock fire involved lead-acid batteries. Most new capacity is lithium-ion, which is a different problem entirely. Lithium cells fail through thermal runaway: one cell overheats — from a manufacturing defect, overcharging, water intrusion, physical damage, even a coding error or a deliberate cyberattack — and it heats the cells next to it, which heat the cells next to those. In a sealed building, the vented gases can build up to the point of explosion. And extinguishing the visible flame doesn’t mean the reaction has stopped: crews see fires “go out” and reignite hours later because the heat inside the module never actually dropped.
Then there’s the water math. A data center fire can take 1 to 2 million gallons of water to extinguish. Round Rock has a municipal hydrant system that can supply that. Most of the towns where data centers are actually being sited — rural Texas, rural Pennsylvania, the four-hours-outside-New-Orleans sites Gary references — don’t. A rural water tender truck carries 2,000 to 3,000 gallons. Getting a million gallons to a fire with a 2,000-gallon truck means roughly 500 round trips between the water source and the fire. That’s the arithmetic a volunteer fire department is looking at when a facility the size of a small city catches fire in their jurisdiction.
The standard that exists — and isn’t being enforced
UL 9540 is the regulation for how battery energy storage systems should be built and separated to contain a fire if one starts. UL 9540A is the test standard that verifies compliance. It was updated in March 2026. Round Rock, which only adopted the 2024 International Fire Code in December, hasn’t incorporated the update yet — and Gary is careful to note this isn’t a Round Rock-specific failure. Most jurisdictions are in the same position, and many rural counties in Texas can’t even have a fire marshal until they hit a population of 250,000. Out of 254 counties, most never will.
Layered on top of outdated code: NDAs. Data center operators frequently won’t share emergency plans, building layouts, or battery composition with the fire departments that would have to respond to their fires. “If there’s something they want to hide that badly,” Gary says, “then maybe they should take their business elsewhere.”
Why this might be the argument that actually lands
Gary’s read, after four appearances before Texas House and Senate committees in the past two months: noise and water complaints get pre-empted. Developers show up to a town a year before residents hear anything, work the local elected officials, and have counterarguments ready by the time anyone objects publicly. Fire risk is new enough that it hasn’t been pre-argued yet — which is part of why his first committee testimony picked up tens of thousands of views after Alec Baldwin reposted it, despite Gary having no public platform of his own.
It also cuts across party lines in a way almost nothing else does right now. Gary points to Texas Governor Abbott’s abrupt shift from “we’re going to be the number one data center state” to ordering an audit — which Gary reads more as pre-election positioning than genuine reform, timed to clear the way for large, established operators (Google, Amazon, Vantage) while filtering out speculative developers who aren’t shovel-ready yet. Even so, he sees it as evidence that the politics have shifted: “opposition to data centers is probably the most bipartisan issue I’ve seen in all my years of voting.”
If you want to do something with this
Gary’s suggestion isn’t complicated. Go to your fire marshal, or your fire chief if there’s no marshal, and ask:
* What’s our current fire code, and does it specifically address battery energy storage systems?
* Has this facility been tested to UL 9540A?
* What training have our firefighters received on battery and electrical fires?
* How many stations, personnel, and apparatus do we actually have, relative to a facility this size?
None of these require expertise. They just require someone to ask, on the record, before the next facility gets approved — because once it’s built, Gary’s experience says getting the answers gets a lot harder.
Protect Round Rock has the fire report and hazmat documentation from the Sabey fire posted at protectroundrock.org, for anyone building the same case in their own town.
What to do next
The full conversation also gets into where this connects to the broader AI risk conversation Gary’s been pulled into by proximity — job displacement, Flock camera surveillance, and the extinction-risk debate — and why he thinks the anti-data-center coalition, being genuinely bipartisan, might be the most persuadable audience for those harder conversations. Worth the full listen if you’re tracking either fight.


This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theairisknetwork.substack.com/subscribe