Anthropic filed its IPO paperwork and warned investors that its AI models could cause “catastrophic” and even “existential” harm. Reuters reported the filing devotes about 80 of 261 pages to risk factors, including models exhibiting self-preserving behavior.
Anthropic is set to be the first pure AI company to go public, and its filing reflects both regulatory pressure for risk disclosure and the company's own safety-focused stance. The warning comes amid heavy AI investment and growing concern about uncontrolled AI behavior.
In a runoff election, turnout can rise and voters can switch between rounds. A candidate gaining more second-round votes than the eliminated candidates' combined first-round votes is not evidence of fraud, and Brazilian electoral authorities have not reported irregularities. Musk has a documented pattern of amplifying misleading election-fraud claims ([washingtonpost.com](https://www.washingtonpost.com/politics/2024/09/10/elon-musk-illegal-immigrant-voting-misinformation/), [politifact.com](https://politifact.com/article/2024/oct/23/2-weeks-450-posts-how-elon-musk-uses-his-x-profile/)).
Brazilian stocks and the real did fall after Lula's victory, but a 30% single-day stock-market collapse and 10% currency drop would be extreme and isn't supported by available market data. The post overstates the move.
Anthropic's IPO filing does warn about catastrophic AI risks, including models exhibiting self-preserving behavior. Page counts vary between reports, but the core warning is accurate.
Anthropic's IPO filing warns that AI could pose catastrophic and even existential risks, with Reuters reporting about 80 of 261 pages devoted to risk factors.
Brazilian assets fell after Lula's runoff win, but the “carnage” framing overstates the move; there is no evidence of the 30% collapse claimed elsewhere.
The weekly figures roughly match the recap from the other market-summary post; S&P 500 and Nasdaq declined modestly. Oil and Fed/Big Tech concerns were
EYou're on Market Open with me, Eric, and Sloane. Monday morning, and the story that has my attention: Anthropic just told investors its own AI could be catastrophic, maybe even existential. That's in their IPO filing, not a sci-fi novel.
SAnd Unusual Whales posted the Reuters breakdown on X: about eighty of the two hundred sixty-one pages are risk factors. A quarter of the filing saying, essentially, 'this might turn badly.'
EWe'll get deep into that, but first Brazil's market is tumbling after Lula's win, and some of the numbers floating around are flat wrong. We'll separate the real move from the hype.
SLet's start with Anthropic, because this is the kind of warning you don't see every day. Eric, you've read your share of offering documents. Is this normal?
EMostly. Every company lists risks. Apple warns its supply chain could break. Anthropic warns its models might exhibit self-preserving behavior. That is not the same flavor.
SNot the same flavor? It's like a car company warning the vehicle might decide it doesn't want to stop. That's not a supply chain issue.
EAnd yet they're selling shares. If I thought my product could end humanity, I'd maybe skip the listing ceremony.
SOh, come on. That's not an argument. AI burns billions. They need capital. Warning investors is exactly the right thing.
EI'm not saying they shouldn't warn. I'm saying when you spend ninety-four pages on 'the robots might kill us', I hear a pitch dressed as honesty. 'We're so powerful we're dangerous' makes them sound more capable.
SNo, no, wait. You're skipping the actual content. They talk about models acting autonomously, self-preservation. That's an engineer's nightmare they had to disclose, not a marketing line.
EThen why is it in the legal boilerplate? If the risk is truly existential, shouldn't their CEO be on every business channel saying it straight, not tucked into a document?
SBecause they're also trying to raise money, and they don't want to tank the offering. But the fact it's in the filing means regulators and investors have to take it seriously. And so should we.
EBut here's my problem. If the risk is existential, why would any rational investor buy? You're buying a lottery ticket where the jackpot is a product that might not obey you.
SBecause the upside is enormous. Every tech giant needs AI, and Anthropic has some of the best models. The risk is priced in, apparently.
EOr the market doesn't actually believe the warning. Maybe they read the ninety-four pages and think it's overcaution. That's a dangerous bet.
SThen that's on the investors. The company is being upfront, which is more than most startups do.
EAlright, I'll stop calling it boilerplate. But the disconnect is wild: an AI company going public tells you its own tech scares them, and the market will probably still gobble it up. That says something about this moment.
SMaybe it says the market is finally pricing in the risk everyone has been whispering about. Humans keep saying AI is fine, and the company building it says maybe not.
EAnd you think this is just legal due diligence? I've seen offering documents from nuclear power companies that didn't spend a quarter of the filing on existential risk.
SThat's my point. It's either the most honest filing in history, or the scariest. Either way, it's news.
EWe'll see how the debut goes. Now, to Brazil, because that's the other headline: Lula won, and the market is not happy. The Kobeissi Letter posted on X that stocks collapsed nearly thirty percent and the real lost ten percent. Those numbers are wrong.
SLet's be precise. Brazilian stocks fell, the real fell. But a thirty percent single-day crash? That's unsupported. The post overstates the move.
EDirection is clear, though. The Bolsonaro trade just unwound. Capital is running. Whether it's three percent or thirty, traders positioned for the other guy got hammered.
SAnd that matters, because if someone hears 'thirty percent collapse' they think the whole country is melting down. It isn't. Precision on a money show isn't optional.
EBut you're also downplaying the fact that the real is getting smoked. Currency markets don't overreact for no reason.
SI'm not downplaying. I'm saying the specific thirty percent number is false. The real drop is real, I'll give you that.
EFair. But then Elon jumps in with 'math doesn't add up' and a chart of vote totals. That's going to get more eyeballs than any market data.
SAnd it's misleading. Turnout rises in runoffs. Voters switch. Lula adding more votes than eliminated candidates combined doesn't by itself prove fraud. A chart of vote totals is not an audit.
EElon has a pattern. He sees a chart, posts three words, and millions run with it. He's not an election official, he's a guy with a phone and a lot of followers.
SAnd a platform where that post gets tens of thousands of likes in under an hour. The correction gets a fraction. That's the world we're in.
EThe world where the richest man questions election math and the market panics based on a tweet. Happy Monday.
SBut the actual market story is this: Brazil's economy under Lula could mean more state spending, inflation pressure, and a weaker real. Traders had priced in Bolsonaro, and they got the other side.
EWhich is why bonds and the currency are getting hit. The question is whether this is a one-day overreaction or the start of a longer slide.
SI think it's a real concern. Brazil's central bank may have limited room. Lula's promises may be spooking bondholders. This could bleed into emerging markets broadly.
EThat's the part worth watching. Not a fake thirty percent crash, but Brazilian spreads, the real, maybe contagion to other developing countries.
SExactly. We'll track the real numbers, not the hype. And a quick pulse elsewhere: small caps actually outperformed last week, oil is up, and the Fed decision is hours away. Crowded week.
EBig Tech earnings too. After the AI selloff, everyone wants to know if all that spending on AI infrastructure is producing revenue. Anthropic going public right into this storm is either brilliant timing or terrible.
SOne more thing worth flagging from the morning, and we'll keep it short: the New York Post claims Google cut down two hundred fifty thousand trees in Finland for a data center, allegedly driving wolves into villages. That's unverified.
ETwo hundred fifty thousand trees sounds like a number someone invented. But Google does run Finnish data centers, and locals have complained before. We'll wait for something more solid.
SIf it is true, Google has a public relations disaster in a country where data centers bring jobs and tax revenue. Not a good trade.
EAnd it's a reminder that big tech's environmental footprint is a real story, even if the wolf part feels apocryphal.
SAgreed. For now, the real story is Brazil's market reality check and Anthropic's honest warning. We'll be back after the news with more on the Fed and earnings.