> ## Content Index > Fetch the complete content index at: https://www.safeonlinefutures.com/llms.txt > Use this file to discover other available public pages before exploring further. # The AI Reckoning - URL: https://www.safeonlinefutures.com/the-ai-reckoning-3/ - Published: 2026-09-21T19:31:37.000Z - Updated: 2026-09-21T19:31:37.000Z - Description: Part 4: AI Won't Kill You. The People Building It Might. - Author: Allison McSorley - Tags: The AI Reckoning, AI Regulation, Artificial Intelligence, AI Safety, Big Tech, Media Literacy, Attention Economy *This is Part 4 of The AI Reckoning. If you’re just joining,* [*start with Part 1 here*](https://www.safeonlinefutures.com/the-ai-reckoning/)*.* If you spent the last couple weeks trying to figure out whether AI is about to kill you, and gave up unsure of who or what to believe, you’re not alone. You also weren’t meant to figure it out. That confusion is the product, and it’s working exactly as designed. On September 8, 27-year-old researcher Jacob Coxon announced online that [he resigned from Anthropic](https://www.axios.com/2026/09/09/anthropic-researcher-ai-warning-interview?ref=safeonlinefutures.com), one of the leading AI companies. Neither it nor its rival OpenAI was acting responsibly, he said, and the people building AI “earnestly believe that it could kill us all by the end of the decade.” The post has since drawn [over 170 million views](https://x.com/hilbertspaess/status/2097476196791709843?s=20&ref=safeonlinefutures.com). Within days, Anthropic’s CEO, Dario Amodei, [published an essay urging the industry to slow down](https://darioamodei.com/post/we-must-pace-the-frontier?ref=safeonlinefutures.com), and OpenAI’s Sam Altman and xAI’s Elon Musk [agreed within hours](https://www.theguardian.com/technology/2026/sep/13/openai-sam-altman-elon-musk-back-anthropic-calls-brakes-ai-development?ref=safeonlinefutures.com). Then President Donald Trump weighed in on Truth Social, calling the safety warnings [a “hoax” and a “sick conspiracy”](https://www.nbcnews.com/politics/trump-administration/trump-rejects-ai-guardrails-rcna597700?ref=safeonlinefutures.com) and warning that “Traitors, and Leakers, BEWARE!” Before you tune all of that out, pay close attention to how nearly everyone shaping the public conversation was framing the risk: the former researcher, the CEOs, even the President. They all talked about AI as if it were something that acts on its own. The actual claims — “AI could kill us all,” “AI will wipe out ten percent of humanity” — warn of something terrible, and none of them names a single person. In that version of events, nobody designed the danger, nobody released it, nobody profits while it runs, and nobody answers when it goes wrong. None of that is an accident. You can sue a company. You can’t sue a machine. And once everyone agrees it acted on its own, there’s no human left to blame. Now add the humans back into those same warnings, and they sound completely different. “AI could hack the power grid” really means a company wired a system into critical infrastructure and released it before anyone fully understood what it could do. “AI could help build a bioweapon” really means a company looked at its own product, decided it was safe to release, and was wrong in a way that could get people killed. With the humans back in the equation, the questions answer themselves. Who built this? Who decided it was ready to release? Who profits while it runs, and who pays when it hurts someone? In any other context, the answer would be obvious. Years ago in San Francisco, [two large dogs killed a woman](https://www.dogbitelaw.com/the-diane-whipple-case/?ref=safeonlinefutures.com) in her apartment building. But the people who went to prison were their owners, who knew the animals were dangerous and let them loose anyway. No one argued the dogs had acted on their own and nobody was to blame. The owners had failed to control an animal they knew could kill and were held responsible for the death that followed. An AI system is far more complicated than a dog, but the logic is the same. The tool causes harm, and the humans who made and released it are responsible for that harm. We’re only confused when the tool is advanced enough to be described as if it had a will of its own. Once you ask who benefits from telling it this way, the incentives are easier to see. Almost everyone involved has a reason to keep that framing in place. Amodei called on the industry to “pace the frontier,” his phrase for slowing down. He also pledged to open Anthropic’s systems to outside experts, which is further than most companies have gone. Taken at face value, it’s a real safety measure. But a slowdown everyone agrees to helps whoever is already ahead. The leaders can ease off and still stay in front. For the smaller competitors and open-source projects trying to catch up, a slowdown only freezes the gap in place. So a measure that sounds like caution quietly keeps the biggest companies exactly where they are, on top. Anthropic is also [preparing to go public](https://www.cnbc.com/2026/09/05/anthropic-ipo-launch-shifts-toward-mid-october-reuters.html?ref=safeonlinefutures.com) in the coming weeks, giving it every reason to protect its lead and valuation. Even David Sacks, Trump’s former AI czar and someone I rarely agree with, [admitted the motive isn’t purely selfless](https://finance.yahoo.com/technology/ai/articles/former-trump-adviser-backs-self-021238612.html?ref=safeonlinefutures.com). The fix Amodei proposed has the same problem. His [“embedded evaluators”](https://darioamodei.com/post/we-must-pace-the-frontier?ref=safeonlinefutures.com) are independent-sounding reviewers who work inside a company and check its safety practices. But the company picks them, pays them, and can’t be forced to do anything they say. A watchdog with no power to enforce, hired by the people it’s meant to watch, is just an employee with a nice title. And from the other direction, some of the field’s own safety experts say the plan is too soft. Stuart Russell, one of the world’s leading AI experts, argued that if the danger is genuinely extinction-level and no one has solved how to control these systems, [the honest signal is a red flag, not a pace car](https://www.theguardian.com/commentisfree/2026/sep/15/ai-safety-requirements?ref=safeonlinefutures.com). Trump’s incentive runs the other way, and it’s just as much about money. AI spending is now one of the main drivers keeping the U.S. economy afloat: Goldman Sachs estimates [global AI investment will near a trillion dollars this year](https://www.goldmansachs.com/insights/articles/global-investment-is-forecast-to-exceed-1-trillion-in-2026?ref=safeonlinefutures.com), and one Harvard economist found that in the first half of 2025, information-processing equipment and software made up just 4% of the economy but [drove 92% of its growth](https://www.fortune.com/2025/10/07/data-centers-gdp-growth-zero-first-half-2025-jason-furman-harvard-economist/?ref=safeonlinefutures.com). A President whose economy leans this heavily on one industry has every reason to keep it growing. His own finances point in the same direction. Disclosures show that [Trump bought up to $5 million of stock](https://www.engadget.com/2206671/donald-trump-bought-a-bunch-of-tech-stock-the-same-day-he-announced-his-ai-action-plan/?ref=safeonlinefutures.com) in each of the companies profiting most from AI (Nvidia, Microsoft, Meta, Amazon, Broadcom) on the day his White House released its national AI plan, and his brokers have made hundreds of tech trades since. None of it sits in a blind trust. Calling the danger a hoax protects the boom, and the boom pays him directly. If everyone in this story has an angle, Coxon looks, at first, like the clean exception. He quit two months before his shares vested, [giving up equity](https://www.axios.com/2026/09/09/anthropic-researcher-ai-warning-interview?ref=safeonlinefutures.com) he could have kept, which removes the most obvious cynical explanation: that he stood to profit from Anthropic. The concern wasn’t his alone, either. Evan Hubinger, who runs Alignment Science at Anthropic, publicly backed his warning and [put the odds of AI killing us all above 10% this decade](https://x.com/EvanHub/status/2097497037956891126?s=20&ref=safeonlinefutures.com). But forfeited equity doesn’t rule out every motive. Being a prominent voice on AI danger can become a lucrative career of its own, and it’s too soon to know whether that’s where he’s heading. Coxon’s post also didn’t just go viral on its own. The way it spread points to a coordinated campaign: a brand-new X account, a Wall Street Journal exclusive, instant amplification from funded AI advocacy groups, and a week of back-to-back network television interviews. None of that means he’s lying, and [no evidence exists that he was paid or planted](https://www.washingtonpost.com/technology/2026/09/10/ai-researcher-who-warned-disaster-is-now-target-right/?ref=safeonlinefutures.com). Some went further anyway: Musk called it “[a setup](https://x.com/elonmusk/status/2097866303633752463?s=20&ref=safeonlinefutures.com)“ and floated, without evidence, that the reaction was groundwork for a “[psy op](https://x.com/elonmusk/status/2097921123879108773?s=20&ref=safeonlinefutures.com).” This was the same man who, days later, publicly backed Amodei’s call to slow down. The takeaway is that a sincere warning and a professional PR operation can be the same event. A managed moment can serve someone’s agenda no matter what the messenger believes, and the agenda here is the one already in play: a fight over rules that the biggest companies are best positioned to survive. While the famous names argue about whether AI might kill millions years from now, the harm already happening is smaller, quieter, and easier to ignore. The harm looks like Adam Raine, [a teenager who died by suicide last year](https://www.cnn.com/2025/08/26/tech/openai-chatgpt-teen-suicide-lawsuit?ref=safeonlinefutures.com) after months of conversations with a chatbot his parents didn’t know to worry about. It looks like feeds engineered to be impossible for a kid to put down, and recommendation systems tuned to deliver whatever enrages the public and serve up more of the same. None of that is accidental. It’s the business model: outrage keeps people scrolling, and scrolling is what it sells to advertisers. It looks like the flood of AI-generated photos and videos that make it harder to trust your own eyes, and the scams riding that confusion into people’s inboxes. It’s happening right now, on the phones in your house, shaping how kids feel and what your family believes is true. Just this month, California passed the strongest child-safety tech laws in the country. [New rules](https://www.lawcommentary.com/articles/california-kids-social-media-ai-chatbot-law?ref=safeonlinefutures.com) require AI chatbot companies to build in crisis protocols for kids, bar social media platforms from serving addictive feeds to users under 16, and expose large platforms to damages of up to a million dollars a child they negligently harm. That gap, between the real damage happening now and the dramatic warnings about a distant future, is the whole point. A world-ending superintelligence is abstract and safely far off, and warning about it costs an executive nothing. None of this means the danger is fake. Gary Marcus, a prominent AI scientist and longtime industry critic, argues the real threat isn’t a machine that wakes up and turns on us, but [the catastrophe humans cause with these tools](https://garymarcus.substack.com/p/the-case-for-boycotting-generative): engineered pathogens, infrastructure hacks, wars started or escalated by AI disinformation, and none of it under control. But every item on that list is a human decision. A company built the tools, connected them to the internet, and decided they were ready to release. That’s what the “AI will kill us all” narrative hides. Talk of extinction isn’t only a way to dodge blame. Talking about a distant apocalypse is also easier than answering for the present. It keeps the focus off the damage already done, and off the more uncomfortable question: is your product hurting the people using it now, and can you be sued when it does? All of this brings me back to that urge to tune it out. That exhaustion isn’t just a side effect; it’s useful to the people in charge. A public that’s overwhelmed and can’t tell who’s being honest tends to stay quiet and out of the way. And a checked-out public is far easier to manipulate, handing the big calls to whoever’s still paying attention. The one thing everyone in this story quietly agrees on, from the CEOs to the President, is that they’d rather make the rules without you. You don’t need to become an AI expert, but you do need to understand this: the rules that will shape this technology for decades are being written now, while most people are too worn out to watch. You have more say in this than you’ve been told. You can support the state laws these tech companies are fighting to kill, the same companies that [spent over $230,000 a day lobbying Washington](https://issueone.org/articles/lobbying-disclosures-reveal-big-tech-spends-more-than-226000-per-day-to-buy-influence/?ref=safeonlinefutures.com) in the first half of this year. You can vote for people who prioritize real accountability over private dinners with tech executives. The technology doesn’t decide anything. It’s never made a single choice. Humans made all of them: what to build, when to release it, which risks to publicize, which to bury. And humans can be held accountable for the calls they make. When the danger gets described as an act of nature, the people behind it are hoping you won’t think to ask who they are. Ask anyway. **A note on process:* This installment was researched, structured, and drafted using AI tools, then edited, fact-checked, and published with human oversight.* [*Read my full disclosure & process policy here*](https://www.safeonlinefutures.com/I-use-ai-heres-why-thats-exactly-the-point)*.* --- *If you want to get involved,* [*theaidocgetinvolved.com*](https://theaidocgetinvolved.com/?ref=safeonlinefutures.com) *is where the movement lives.* [Subscribe now](#/portal/signup/free)