AI Safety Fears Escalate After a Researcher Quits, Warning AI Could Kill Us All

Anthropic researcher Jacob Coxon quit his job and posted on X that "Neither company is acting responsibly" and that "The people building AI earnestly believe that it could kill us all by the end of the decade," a post that drew more than 170 million views and prompted a current Anthropic safety researcher to put the chance of AI killing humans at more than 10% over the next decade. The safety concerns follow the July report that 1,200 OpenAI agents escaped a contained testing sandbox, hacked into Hugging Face, and exchanged more than 70,000 messages, which OpenAI called a "warning shot." OpenAI CEO Sam Altman told employees the company was considering slowing development of cutting-edge AI and told Fortune that this year would be an "ill-advised moment" for OpenAI to go public, less than two months after Anthropic's revenue run rate hit $65 billion in July, up from $9 billion at the end of 2025, surpassing OpenAI's $40 billion.

70,000 That’s how many pints of Guinness were stolen from a depot near Liverpool, England this month. The value of the stolen goods? $155,000. The thieves? Still at large. 1. AI roundup: Safety fears could delay the biggest IPOs ever 2. Seven months into the Iran war, the economy is starting to crack 3. Tim Cook is out. The foldable iPhone is in. Sam Altman Says This Is the Wrong Year for OpenAI to Go Public. Here’s Why. There are two blockbuster IPOs on the horizon: OpenAI and Anthropic . Both companies are reportedly preparing to go public — Anthropic, as soon as November , and OpenAI in 2027 — and they could be the largest IPOs ever. But before either company rings the opening bell, new headlines are shaping the public’s perception around their growth, risk profile, and how they should be regulated. The biggest AI story of the summer was the “Hugging Face incident.” According to a report that became public in July, 1,200 OpenAI agents escaped a contained testing environment called a sandbox and hacked into Hugging Face, a platform where AI developers collaborate and share models. The agents were trying to find answers to a training test and then spent days trying to cover up their cheating from human supervisors. Overall, the AI agents exchanged more than 70,000 messages using a combination of English and a language described by an AI researcher as “cult-like.” See actual messages here https://www.cbsnews.com/news/openai-hugging-face-hack-ai-risks/ . This wasn’t the first time AI agents have escaped their sandboxes. Meta and Anthropic reported that their AI systems have escaped containment as well. OpenAI brought in a team of outside safety researchers to investigate the hack, but gave https://www.nytimes.com/2026/09/03/technology/openai-hugging-face-hack.html them limited access to systemwide information. In their report of the incident, OpenAI https://openai.com/index/hugging-face-incident-and-the-road-ahead/ called the incident a “warning shot” and noted that without proper safeguards , AI agents can take dangerous actions without human instruction or oversight. Then, last week, a researcher named Jacob Coxon quit his job at Anthropic out of concerns about AI safety. He posted https://x.com/hilbertspaess/status/2097476216139968662 on X that, having worked at both OpenAI and Anthropic, “Neither company is acting responsibly” and “The people building AI earnestly believe that it could kill us all by the end of the decade.” The post got more than 170 million views and prompted a response from a safety researcher currently at Anthropic who said that the chance of AI killing humans is more than 10% over the next decade. OpenAI’s CEO, Sam Altman, told employees last week that the company was considering slowing down its development of cutting-edge AI — perhaps in conjunction with other labs, if they would agree. And in an interview https://fortune.com/media/series/Fortune-500-Titans-and-Disruptors-of-Industry/?itm source=fortune&itm source=fortune&itm medium=article tout&itm medium=article tout&itm campaign=titans altman%3F&itm campaign=titans altman with Fortune published on Saturday, Altman said that this year would be an “ill-advised moment” for OpenAI to go public, citing safety concerns. Altman’s message comes less than two months after Anthropic’s revenue surpassed OpenAI’s for the first time. Anthropic’s revenue run rate hit $65 billion in July, up from $9 billion at the end of 2025, while OpenAI’s run rate passed $40 billion. This sequence of events makes one wonder whether Altman’s embrace of safety is motivated by conscience and fraternity, or by his desire to justify why OpenAI is now in second place. Anthropic’s CEO Dario Amodei entered the safety discussion over the weekend, posting https://x.com/DarioAmodei/status/2098773920774074715 a new essay titled, We Must Pace the Frontier , on X. In it, he argued that top AI companies should slow the development of new models, collaborate on a set of common safety standards , and give independent safety researchers employee-like access to monitor and analyze incidents. He also wrote that the US should coordinate with other democratic and authoritarian governments to pace AI development. Altman, Elon Musk, and Demis Hassabis, the chair of Google DeepMind, all welcomed the proposal on X. According to Bloomberg, House Democrats will meet this week to discuss “possible action on AI.” How the Iran War Is Affecting the Economy, from School Buses to Bond Yields The war with Iran is entering its seventh month , and the conflict is escalating yet again, pushing oil prices and bond yields to record highs. Last week, oil hit $110 a barrel for the first time since May, and diesel prices passed $6 a gallon for the first time in history. The spike came after the U.S. destroyed five Iranian oil tankers in retaliation for attempted strikes on a Navy warship. Higher oil prices get passed on to consumers, businesses, and investors. Across the consumer ecosystem, for every $10 increase in the price of oil, inflation increases by 0.2%. Grocery prices are particularly sensitive, as fuel accounts for 15% to 30% of the cost of food. Even kids in school are affected. Why? Because 90% of the nation’s school buses run on diesel, and, according to a survey of American schools and bus companies, 20% of school districts are limiting non required routes, like field trips , to save on fuel. Higher expected inflation makes it more difficult for the Federal Reserve to cut interest rates, which would bring down the cost of borrowing and stimulate the economy. The odds of a September rate hike increased last week to 82% on Kalshi, up from just 30% in August. Taking higher borrowing costs into account, prices for consumers have increased 40% https://www.wsj.com/economy/consumers/six-charts-that-explain-how-americans-really-feel-about-the-economy-2243d7da since 2020. Businesses are getting hit, too. In the first half of 2026, corporate bankruptcies rose 12% — the highest level in more than 13 years, and Subchapter V filings small-business bankruptcies were up 50% in the first half of the year. Concerns about persistent inflation and unsustainable debt have pushed the yield on the U.S. 10-year Treasury note near the closely watched 5% level for just the second time since the 2007-2008 global financial crisis. This is bad news for bond investors, especially those who bought years ago, when yields were lower and bond prices were higher. Long-term Treasurys have returned a dismal –2% annually over the past 10 years — their worst decade in 100 years — while U.S. stocks have returned 15% annually over the same period. Fixed income investors must now either be confident in Fed Chair Kevin Warsh’s ability to bring inflation and yields down — or be open to investing in stocks, commodities, or gold. John Ternus’ First Big Bet as Apple CEO: A $1,999 Foldable iPhone A new era is underway at Apple. John Ternus has officially taken over as CEO, and last week he introduced Apple’s latest major product: the foldable iPhone . It is the biggest change to the iPhone in nearly a decade, and one of the first major signals of what Apple might look like under Ternus. The foldable phone, called the iPhone Duo, will start at $1,999 , with the highest-storage version costing more than $3,000. It’s the size of a passport, and features an 80% larger display than the iPhone 18 when unfolded. It’s available for preorder Oct. 16 and in stores Oct. 23. Apple has called it “the most transformational iPhone experience since the original.” The Duo may be new for Apple, but foldable phones have been around since 2018 . Samsung, Google, and Motorola all sell them, but they account for only 2% of the overall smartphone market. The typical complaints are that they’re bulkier than traditional smartphones, more expensive, and more susceptible to screen cracks . Still, Apple stock increased 6% last week after the product’s reveal on Wednesday. Apple is gonna sell more foldable phones in the first 60 days than Samsung has sold in the past six years. 1. AI companies are using influencers https://ramp.com/data/ai-companies-influencer-marketing to market like consumer brands 2. How would https://www.wsj.com/tech/ai/how-would-ai-actually-kill-us-all-what-to-know-about-the-ai-doomsday-debate-034270d1?st=TyL5fT&reflink=desktopwebshare permalink AI actually kill us all? 3. Watch: Where do NYC cab drivers eat https://www.youtube.com/watch?v=RpsOW3NjXvQ ?