Sunday’s tech news November 8 2026 is a study in contrasts: the same AI tools getting more powerful are also getting more dangerous to use carelessly, the companies behind them are pouring billions into the physical infrastructure to run them, and the price of using them keeps falling. Here’s what mattered this weekend.
Key Takeaways
- A convenience click in Claude Code saved passwords and API keys into a file that ended up public on GitHub — 1,122 files and counting.
- Google took over a troubled Wyoming data center project from Crusoe and plans a $17 billion campus there.
- ChatGPT can now turn images into animated GIFs with Images 2.6, and it generates images 50% faster.
- Claude can now read your bank statements — but it can’t move a dollar.
- AI token prices fell nearly 27% in a month as OpenAI and Anthropic undercut each other.
1. One Click in Claude Code Put Passwords on GitHub
The scariest story of the weekend starts with a feature designed to save developers time. According to security firm Apiiro, Claude Code’s “Yes, and don’t ask again” button can quietly save a password or API key into a project settings file — in plain text — and that file has been turning up in public GitHub repositories.
Here’s how it works. When Claude Code wants to run a shell command, it asks first. Pick “don’t ask again” and it stores the exact command as a permanent allow rule in .claude/settings.local.json inside the project folder. If that command happened to include a database password or a cloud key typed inline, the secret goes in with it. The file is supposed to stay on the developer’s own machine, and Claude Code tries to keep it out of version control — but only by adding it to the user’s personal git excludes, not the project’s own .gitignore, so the protection doesn’t travel with the repo. Standard secret scanners missed about one in five of the leaking files, since a password typed into a command doesn’t look like a formatted API key, and in half the affected repos the file had sat there for at least 217 days. Apiiro says developers have flagged this in bug reports since August 2025 with no reply from Anthropic so far. If you use Claude Code, rotate every credential the file has ever held first, then clean it out of your repos and git history.
Apiiro scanned public GitHub between November 1 and 3 and found 104,185 copies of the file across 103,617 repositories. Of those, 1,122 contained at least one real secret — 1,379 distinct secrets in total, including 379 database passwords, 122 Supabase service keys, 43 SSH logins, and 23 AWS keys. Counting older versions in git history, the shared settings file teams commit on purpose, and saved agent chat logs, the total reaches 2,539 secrets across 1,994 repositories, touching 63 organizations. One company confirmed to the researchers that its leaked key had been live.
2. Google Takes Over Crusoe’s Wyoming Data Center — and Pledges $17 Billion
Google is going big in Wyoming. According to TechCrunch, Crusoe confirmed on Friday that it has sold its unfinished data center project in southeastern Wyoming to Google, which will take over development of the site.
The handover follows Google’s own announcement a day earlier of a $17 billion investment in its first Wyoming data center, a campus in Laramie County. The power plan leans on a 702-megawatt fuel cell plant built with Bloom Energy, plus gas generation and grid power, with carbon capture planned down the line. Google says the facility will use air cooling to limit local water use, and expects it to generate hundreds of millions of dollars in property taxes for the county.
The backstory is a little awkward for Crusoe. The Denver-based developer, recently valued at $30.9 billion, originally announced the campus with Blackstone-backed Tallgrass Energy as a 1.8-gigawatt site that could eventually scale to 10 gigawatts. It paused development in June after Google — which it had tried and failed to land as a customer — raised concerns about cost and construction timelines. Now Google owns the site outright, while Crusoe keeps developing a separate Google campus near Amarillo, Texas.
3. ChatGPT Images 2.6: GIFs, Faster Generation, and Better Edits
OpenAI gave ChatGPT’s image tools their biggest upgrade in months. According to OpenAI, Images 2.6 launched on November 4 across ChatGPT, Codex, and the API — and the headline feature is one users have wanted for years: animated GIFs.
You can now turn a prompt or a photo into a looping GIF, adjust the playback speed, and edit individual frames before sharing. The update also adds reusable reference images, so a character, pet, or product stays consistent across new creations without re-uploading it every time. For developers, two new API models arrive: GPT-Image-2.6 Flare, aimed at speed (as little as six seconds per image at low quality), and Sunburst, aimed at higher quality and stronger consistency across edits. Image markup and SVG exports round out the release.
OpenAI says Images 2.6 generates images 50% faster than before, and it’s available globally across all subscription tiers, including Free. The scale behind it is staggering: people now create 4 billion images a week across ChatGPT, Codex, and the API.
4. Claude Can Now Read Your Bank Account — But It Can’t Touch Your Money
Anthropic is bringing Claude into your finances, carefully. According to Anthropic, Personal Finance in Claude launched in beta on November 5, letting US subscribers link bank, card, loan, and investment accounts through Plaid and ask Claude plain-language questions about their own money.
The key boundary is that access is strictly read-only. As Anthropic puts it, “Claude can’t move, manage, or invest your money.” Claude can see balances, credit limits, and up to two years of transactions for each account you connect — up to 16 banks — and answer questions like where the month’s money went or what’s due before payday. It can’t pay a bill, cancel a subscription, or place a trade, and it isn’t designed to give investment advice.
The rollout is narrow for now: a gradual beta for Pro, Max, and Ultra subscribers in the US only, with no date for other countries. Anthropic also says the feature has no ads or sponsored recommendations, and that chats in the Personal Finance hub aren’t used for model training. It’s the last of the big AI assistants to add account-connected finance — ChatGPT got there first — but the read-only design and the honest caveat that Claude can make mistakes make it a sensible first step.
5. AI Is Getting Cheaper Fast — and That’s Squeezing the AI Labs
Here’s the number that explains a lot of this week’s AI business news: the price of AI tokens is falling off a cliff. According to TechCrunch, new data from spend-management platform Ramp shows businesses used 37.1% more AI tokens in the week ending October 23 than a month earlier — but their spending rose only 8.9%, because the blended price per million tokens fell 26.8% to $0.56.
The price war is reshaping the leaderboard, too. Anthropic’s share of token spending in Ramp’s sample jumped to 60.5% from 51.3% over four weeks, while OpenAI’s fell to 35.1% from 44.2%. OpenAI slashed API prices in half for two GPT-6 models in September; Anthropic answered with cuts of its own. Anthropic even disclosed in a recent filing that price reductions shaved about $600 million off its revenue growth.
For users, this is mostly good news — the same AI costs less every month. For the labs, it’s a squeeze: they have to keep growing usage faster than prices fall, all while funding the data centers from story number two.
The Takeaway
This weekend’s tech news November 8 2026 tells one connected story: AI is getting cheaper, more capable, and more deeply embedded in daily life — and the growing pains are showing up everywhere. Coding agents need better guardrails before they leak your secrets, data centers need new owners when timelines slip, and the labs themselves are racing to the bottom on price while spending billions on the infrastructure underneath it all.
If you missed yesterday’s roundup — the AI stock rebound, OpenAI’s $1.25T valuation, and AI that dreams — catch up there, and check Friday’s crash and OpenAI’s hacker model for the other half of this week’s wild ride.
FAQ
What happened in the Claude Code password leak?
Security firm Apiiro found that Claude Code’s “don’t ask again” option saves approved commands — passwords included — into a settings file that ended up in public GitHub repos. Across 1,122 files, researchers found 1,379 real secrets, and 2,539 in total counting git history and shared settings files.
How much is Google investing in its Wyoming data center?
Google announced a long-term $17 billion investment in its first Wyoming data center, a campus in Laramie County, after acquiring the site from Crusoe. The campus will be powered largely by a planned 702-megawatt fuel cell plant.
Can ChatGPT make GIFs now?
Yes. With Images 2.6, launched November 4, ChatGPT can turn a prompt or image into an animated GIF, with controls for playback speed and individual frames. The update is available on all ChatGPT subscription tiers, including Free.
What is Claude Personal Finance?
It’s a beta feature in Claude that links your US bank and card accounts through Plaid so Claude can answer questions about your balances and spending. Access is read-only — Claude can’t move, manage, or invest your money — and it’s rolling out gradually to Pro, Max, and Ultra subscribers in the US.
Sources
- Apiiro — Claude Code’s “Don’t Ask Again” exposes passwords in 1,122 public GitHub files
- TechCrunch — Crusoe sells troubled Wyoming data center project to Google
- OpenAI — Introducing Images 2.6
- Anthropic — Introducing Personal Finance in Claude
- TechCrunch — Ramp data shows AI spending stalling just as Anthropic nears its IPO
