A brutal day for AI stocks, a report that AI is now speeding up its own development, and a chatbot that can read your bank account — Friday’s tech news in one place.
The Market’s Worst Day in Over a Year
Thursday was the worst day for AI stocks in more than a year, and the numbers are hard to ignore. The Nasdaq Composite fell 5.52%, its steepest one-day drop in nineteen months. The S&P 500 fell 4.36%, and the Dow dropped 3.44%. Roughly $3 trillion of market value was wiped off S&P 500 companies in a single session.
Chip stocks took the worst of it. Nvidia fell about 9.3%, its biggest one-day fall since January 2025. AMD plunged 17.3%. SanDisk sank 22%, and the Philadelphia Semiconductor Index dropped more than 12%. The Nasdaq is now roughly 12% below the record high it set in early October, which puts it in correction territory.
Politics, Not Products, Drove the Selloff
According to Reuters, the trigger was politics, not products. Democrats swept this week’s midterm elections and are expected to take control of Congress, and investors are worried that means tougher regulation of AI, from data center permits to model safety rules. No major AI company announced a cancelled product or a lost customer on Thursday. This was a repricing of risk, not a broken business.
The rebound started almost immediately. On Friday morning, Nasdaq futures were up, Nvidia was up about 3.6% in premarket trading, and AMD, Intel, Broadcom, and Micron were up between 5% and 8%. One strategist at Jefferies called Thursday’s selloff reactionary, not a fundamental reset.
The bigger question is whether the AI buildout itself slows down, if regulation makes data centers harder to build and investors get nervous about the hundreds of billions being spent on chips. For now, the companies are still spending.
AI Is Now Accelerating Its Own Development
An independent research group called METR spent eleven weeks inside Anthropic, with access to internal data, benchmark scores, and staff interviews, to answer one question: is AI speeding up the development of AI itself?
Their answer, published Thursday, is yes, and by a lot. METR estimates that Anthropic’s models have been improving about 1.9 times faster since October 2025 than they were before. In plain terms, Anthropic made roughly two years of historical progress in the last year. METR estimates that perhaps 80% of that acceleration comes from AI itself helping with the research.
There are big caveats. Anthropic redacted a majority of the evidence METR found most informative, saying it could help rival labs. METR says its estimate comes with high uncertainty — the true acceleration could be as low as 1.3 times, or as high as 3 times.
Still, this lands at a sensitive moment. Anthropic’s own IPO filing, amended earlier this week, warns that future models may imminently meet its own threshold for dramatically accelerated progress driven by AI. The company says no current model has crossed that line. METR’s report is the first independent look at whether that line is closer than anyone thought, and it raises the question Anthropic’s chief executive has been asking publicly: if AI keeps accelerating its own development, can safety work keep up?
OpenAI’s First “Critical”-Rated Cyber Model
OpenAI had a big security announcement of its own on Thursday. It released GPT-6 Cyber, a model built specifically to find and exploit security flaws, and it gave the model a Critical rating for cybersecurity under its safety framework. That is the highest rating in the framework, and the first time OpenAI has ever assigned it.
This is the same model behind the Linux kernel research discussed in yesterday’s episode, where an OpenAI pipeline found 948 potential privilege escalation vulnerabilities. OpenAI says GPT-6 Cyber also found 81 previously unknown vulnerabilities in Firefox that Mozilla classified as high severity, and Mozilla is already releasing patches.
On the benchmarks, GPT-6 Cyber scored 48.5% on ExploitGym, ahead of 42.4% for the flagship GPT-6 Astra and 32.1% for the previous cyber model. In a test designed to catch the model wandering outside its authorized scope, it stayed inside the lines more than 99% of the time, down from about 5% of escapes for its predecessor.
Access is tightly controlled, for obvious reasons. The model is going to a small group of vetted customers in OpenAI’s Daybreak Red program, for authorized offensive security work like penetration testing. OpenAI says access will expand to all approved Daybreak Red customers over the coming weeks. It also announced a $1 billion initiative to support cybersecurity professionals working on critical infrastructure, like the power grid and water systems.
Claude Can Now Read Your Bank Account
Anthropic announced Thursday that Claude is getting a new tool called Personal Finance, and it is a significant step into territory where chatbots have not gone before.
Here is how it works. You connect your bank and credit card accounts through Plaid, the same service budgeting apps use, and then you can ask Claude questions about your money in plain language — where did my spending go last month, am I on track with my bills, that kind of thing.
The key detail is that access is read-only. Claude cannot move, manage, or invest your money. It cannot transfer funds, pay a bill, or cancel a subscription. Anthropic published exactly what Claude can see for each account: the bank’s name, the account name and type, the last four digits of the account number, balances, credit limits, and up to two years of transaction history.
It is rolling out in beta this week to paying subscribers in the United States, on the Pro, Max, and Ultra plans. Anthropic says conversations inside the Personal Finance hub will not be used to train its models, and the feature is not designed to give investment advice. ChatGPT added a similar feature earlier this year, so this is Anthropic catching up in a race to become your everyday financial companion.
The trust question is the whole story here. A chatbot that can read two years of your bank transactions is useful only if you believe the boundaries will hold. Read-only, no passwords, revocable access, and no training on your conversations — those are the promises.
Google’s $17 Billion Wyoming Bet — While Voters Push Back
Google announced Thursday that it will invest $17 billion to build its first data center in Wyoming, in partnership with Tallgrass Energy, Black Hills Energy, American Electric Power, and Related Digital. The project will be powered in part by a planned 702-megawatt fuel-cell development with carbon capture, supplied by Bloom Energy.
The timing is striking, because Google is pushing ahead while political resistance to data centers is growing. Of the 41 data center-related ballot measures voted on this week, all but one went against data centers, according to Ballotpedia. The AI buildout needs enormous amounts of power and land, and local communities are increasingly pushing back.
The Through Line: Trust
Thursday was a brutal day for AI stocks, but the selloff was driven by politics and nerves, not by anything breaking in the technology. The technology itself is moving faster than ever, with METR saying AI is now accelerating its own development, and OpenAI shipping a cyber model it rates as critical.
And the AI companies are moving into more of your life, from Claude reading your bank statements to Google building data centers in your backyard. The through line today is trust. Investors are asking whether they can trust the AI boom. Regulators are asking whether they can trust the labs. And users are being asked to trust chatbots with their money.
Sources
- AI stock selloff November 5: Reuters
- METR finds Anthropic AI development accelerating: Reuters
- OpenAI launches GPT-6 Cyber: Reuters
- Claude Personal Finance read-only bank access: TechCrunch
- Google plans $17 billion Wyoming data center: Reuters
Adapted from the November 6, 2026 episode of Daily Tech and AI.
Daily Tech and AI is AI-voiced and produced by Angelo Ruiz.
Leave a Reply