Voters block AI data centers
Voters rejected data center development in 40 of 41 ballot measures across 10 states on Tuesday, delivering a sharp rebuke to the AI infrastructure boom that Big Tech and hyperscalers have been racing to build.
- Ohio accounted for more than half of the measures, with 21 on the ballot, and Democrat Amy Acton flipped the governor’s seat after supporting a “conditional moratorium” requiring data centers to cover 100% of utility costs and use union labor.
- In Texas, Democrat Gina Hinojosa unseated Gov. Greg Abbott by framing the Republican incumbent as prioritizing tech industry interests over local communities, a warning sign for companies that left California seeking friendlier business climates.
- 56% of voters in a CNBC poll said data centers were a crucial or major factor in their decision, with Democrats ranking the issue higher than Republicans and already calling for AI executives to testify under oath.
For enterprises betting on AI infrastructure expansion, the political calculus just shifted. OpenAI’s 8-gigawatt Ohio campus, backed by SoftBank and Nvidia, now faces a potentially hostile state government. The hyperscalers are each spending hundreds of billions annually on data centers, and community resistance will add new hurdles to permitting timelines, utility negotiations, and site selection. Companies planning large-scale AI deployments should factor rising political risk into capacity forecasts.
Apple is cutting AI budgets for internal engineering teams, reversing months of aggressive adoption.
The company lowered its daily AI spending cap for engineering teams from $300 to $200 in October. Further cuts are coming, with some allowances dropping to $100 a day. Non-engineering teams may face limits as low as $25 daily.
The pullback follows a wave of “tokenmaxxing” earlier this year, when companies pushed staff to maximize AI usage in pursuit of efficiency gains. Bills came due. Companies learned that token rationing makes more sense than unlimited access.
Apple’s move signals a broader industry recalibration. The gap between AI operating costs and measurable productivity gains is forcing enterprises to treat AI budgets like any other expense, not a blank check for innovation.
Claude Is Hiding a Secret Mark in Your Text
Since November 3, Anthropic has been adding an invisible watermark to text from older Claude models too, not just its newest ones.
- The mark now covers Claude Mythos 5, Opus 4.7, Opus 4.6 and Sonnet 4.6, according to an email Anthropic sent customers.
- Newer models such as Claude Opus 5.5 have carried it since Anthropic announced the system in August.
- Anthropic says it doesn’t change the meaning, quality or readability of the text, and it survives copy and paste.
- You can’t check it yourself. Detection is a private preview for groups like regulators, journalists and researchers.
- The mark can fade in very short, heavily edited, paraphrased or translated text.
If you write with Claude, a hidden trace may now travel with your words. Even text Claude only proofread or translated can carry it.
So a detected mark shows Claude touched the text, not that Claude wrote it. Expect a lot of arguments about what that means for students and job seekers.
ChatGPT Can Finally Make GIFs
OpenAI launched ChatGPT Images 2.6, which can turn still images into short animated GIFs and remember reference photos of you, your pets or your characters.
- Turn a photo, or even a basic sketch, into a short animation for reactions, memes or mascots.
- Save reference images of people, pets and characters so you don’t have to upload them again.
- OpenAI says it creates images 50% faster than the previous version.
- It’s live in ChatGPT and Codex on web, desktop and mobile.
- Developers get two new API models: GPT-Image-2.6 Flare (as fast as about 6 seconds) and Sunburst (more detail, plus SVG export).
GIFs were one of the last things you still needed a separate app for.
Now memes, reactions and animated logos can come straight out of ChatGPT. Scroll down to the tutorial to make your first one.
Turn Any Photo Into a GIF with ChatGPT
Your pet, your selfie or your product shot, animated in under a minute. No GIF app needed.
Steps to Follow:
- Open ChatGPT on web, desktop or mobile.
- Upload a photo. Pets and faces work great.
- Paste the prompt below and fill in what should move.
- Not quite right? Ask for one small change, like “slower” or “only the tail moves”.
- Download it and use it as a reaction, story or post.
Copy this prompt:
Turn this photo into a short looping GIF. Make [the dog's tail wag], and keep everything else perfectly still.
Claude Code’s Product Lead Joins OpenAI
Cat Wu, who led product for Claude Code and Cowork at Anthropic, is joining OpenAI to lead product management for its Core Products team.
- She joined Anthropic in August 2024 and helped grow Claude Code from a small internal project into one of its best-known products.
- Her last day at Anthropic was October 30.
- At OpenAI she’ll work with Thibault Sottiaux, whose team covers ChatGPT, the API and Codex.
- She says her priority is making AI agents reliable.
- Reports say she’s the third person to leave the Claude Code team in about two weeks.
Claude Code and OpenAI’s Codex are fighting for the same developers, and the person who shaped one now works on the other.
The timing is awkward for Anthropic, which is reportedly preparing to go public.
Isomorphic Labs Eyes a $50B Valuation
Google DeepMind’s drug-design spinout is reportedly in talks to raise about $5B, up from a $2.1B round just five months ago.
China Blocks SoftBank’s $5.4B Robot Deal
Beijing reportedly won’t clear SoftBank’s purchase of ABB’s robotics unit, a blow to Masayoshi Son’s AI push. SoftBank shares fell about 5%.
Claude Finds Critical Flaws in Xen
A new Xen security advisory credits Claude with finding two bugs that let a malicious virtual machine read other machines’ memory.
Ohio Voters Push Back on AI Data Centers
Voters passed data-center bans or restrictions in at least 14 Ohio towns, adding to more than 125 local moratoriums across the US.
Netflix Tests an AI Mood Picker
A new AI “Discover” tab on supported US TVs suggests what to watch based on your mood or who you’re watching with.
OpenAI’s AI Found 948 Ways to Hack Linux
OpenAI pointed an AI security pipeline at the Linux kernel. In one run it produced 948 working exploits that could give an attacker full control of a machine, plus a suggested fix for each.
- The AI flagged 2,308 possible bugs and confirmed 1,724 of them as real.
- 948 were “local privilege escalation” flaws, where a normal user can take over as the all-powerful root admin.
- The whole run cost about $1.2 million in AI usage, roughly $1,270 per bug.
- OpenAI isn’t publishing the exploit code, and it’s working with Linux maintainers to get fixes in.
- Security experts review every finding before it reaches a maintainer. None of the fixes have shipped yet.
Linux runs most of the world’s servers, phones and cloud. AI can now find serious holes in it at a scale human researchers can’t match.
That cuts both ways: defenders can fix bugs faster, but attackers with the same tools can find them faster too. Keep your devices updated.
Riot Leases 910 MW to Anthropic
Riot’s powered-land lease with Anthropic at its Corsicana, Texas, site became binding on November 4. Riot plans to deliver 910 MW of critical IT capacity in three phases from June 2027 to July 2028. Anthropic will pay rent as blocks are delivered and cover electricity and other site expenses.
OpenAI Agrees to Buy Pierre Computer
OpenAI agreed to acquire Pierre Computer, the startup behind open-source developer tools already used in Codex. Its libraries handle tasks including rendering code diffs and highlighting syntax. Terms were not disclosed.
Starlink Taps Qualcomm for 5G
SpaceX selected Qualcomm’s Dragonwing chips and software for Starlink Mobile’s planned nationwide terrestrial 5G network. The ground network would complement Starlink’s direct-to-phone satellite service and give SpaceX a path into the U.S. wireless market.
Google Backs Anthropic’s Kentucky Leases
Google signed three agreements to backstop certain Anthropic obligations at TeraWulf’s Kentucky campus. The agreements take effect as the last data hall under each lease starts service. TeraWulf previously estimated about $19 billion in contracted revenue over the initial 20-year lease term.
Qualcomm Slides on Weak Earnings Outlook
Qualcomm beat fiscal fourth-quarter estimates, but its earnings forecast fell short of Wall Street expectations at the midpoint. Shares fell about 7% in extended trading. Rising memory costs have pushed up handset prices and weighed on smartphone demand.
Toyota and Waymo Plan Japan Robotaxi Tests
Toyota and Waymo said on October 30 that they plan to begin testing and validating autonomous bZ4X electric SUVs in Japan in 2027. They plan to bring the vehicles to the U.K. and other Waymo markets in subsequent years.
Tesla Tests Cybercabs on Detroit Streets
About two dozen Tesla Cybercabs are testing on Detroit-area public roads out of a Southfield facility, The Detroit News reported. The vehicles use temporary steering wheels and human test drivers.
Big Tech Buybacks Fall to Multiyear Low
Combined third-quarter buybacks at Alphabet, Amazon, Microsoft, Meta and Oracle fell to $4.4 billion, the lowest since 2017, according to The Kobeissi Letter. Separately, Reuters found that the same companies returned 9.3% of operating cash flow to shareholders through dividends and buybacks in the first nine months, down from 27.4% a year earlier.
Nvidia May Fill 1% of China RTX Pro 5500 Orders
Nvidia is expected to fill about 1% of roughly six million RTX Pro 5500 orders placed by major Chinese cloud and AI companies, Bloomberg reported. The orders were estimated at about $70 billion. The figure applies to that chip and group of buyers.
Anthropic Uses Claude to Map a Viral Protein Shell
Anthropic researchers used Claude to help reconstruct a vault-like particle made from a bacteriophage protein. Claude assisted with DNA design, purification and microscopy analysis. The team says it obtained an initial 3D model 34 hours after microscopy data processing began. The biological function of vaults remains unknown.
IonQ Closes Acquisition of Elisity
IonQ completed its all-stock acquisition of zero-trust security company Elisity, issuing $1.15 billion in shares at closing. Sellers can earn up to another $150 million based on revenue milestones. The deal adds identity-based network security to IonQ’s quantum security portfolio.
Mistral Publishes Large 4 Weights
Mistral Large 4’s weights are available on Hugging Face. The coding and cybersecurity model has 1 trillion total parameters and 52 billion active parameters. Its custom license requires a separate commercial agreement for businesses that operate model-as-a-service offerings before commercial use.
Anthropic Releases FlashProve Research Code
Anthropic published code for FlashProve, a prototype that uses cryptographic proofs to check that model output came from committed weights. In a batched test with 4-bit Llama 4 Maverick, proof generation reached about 110 tokens per second and required 70–95 times the GPU time of inference alone.
SpaceXSI’s Grok Bot starts making phone calls
SpaceXSI just rolled out phone calls in Grok Bot, the company’s AI agent platform, which can now sit on hold, book appointments, and negotiate bills on U.S. numbers — with users legally responsible for every call the bot makes.
- If a rep asks something the bot can’t answer, users get a message mid-call and type a reply for it to pass along, all while following a live transcript.
- SpaceXSI’s terms make users the “caller of record,” “solely responsible” for calls and for following phone laws, with bots disclosing they’re a “virtual assistant.”
- The calls appear to run directly inside the platform via a temporary number, with early examples including a Comcast bill negotiation and a pizza order.
- Grok Bot is available for paid users, and Elon Musk answered complaints about usage limits by calling the platform’s cost “unsustainably high relative to revenue.”
AI labs keep pitching agents as a way to take annoying life tasks off people’s plates, and for many, phone calls sit near the top of that list. With Grok Voice already answering 20,000 Starlink calls a day by SpaceXSI’s own count, the phone could soon go the way of the agentic internet — bots talking with other bots.
Which free AI tool should you actually use?
ChatGPT, Claude, Copilot, Gemini, Grok, and Perplexity all have free plans. But they offer different tools and limits. In this guide, you’ll run the same tasks in each one, check the results, and build a small AI toolkit without paying for the six.
Step-by-step:
- Open the six services and check that each account is on its free plan. Use personal accounts and keep private work out of the test
- Paste the same fictional Repair Café brief into each assistant (all resources in the full guide). Check the event time, staffing, budget, and missing details
- Ask each tool the same question about Shenandoah National Park. Open the official National Park Service source and check the leash rule and trail restrictions
- Upload the same small budget CSV wherever the free tier allows it. If uploads are unavailable, paste the data and record that limit separately
- Score each result for accuracy, source quality, file handling, and usefulness. Choose a main assistant, a research tool, and a backup based on results
Pro tip: A free plan can be generous for chat but tight on uploads, image creation, or research. Check the limit that applies to the task you actually do.
Midterm voters turn on data centers, AI
Voters in U.S. midterms backed limits on data centers almost everywhere those measures appeared on local ballots and handed Congress to Democrats, promising tougher oversight of Silicon Valley, with a majority also telling NBC they expect AI to do more harm than good.
- Data center opponents have taken 38 of 39 fully counted local measures, according to election trackers, with only Harwood, N.D., voting the other way.
- In NBC’s exit poll, 76% of voters would oppose a data center going up in their own community, and 57% expect AI to hurt society more than help.
- Goldman Sachs reportedly sees 45% of planned U.S. data center capacity at risk of temporary freezes, but figures the political heat will cool over time.
- Elon Musk, the biggest midterm donor at $446M backing Republicans, posted that data centers in space look “pretty appealing right now,” following the results.
The public has heavily rebuked AI and the data center buildout in the midterms. With AI infrastructure hubs like Texas flipping blue and Democrats about to run the hearings on AI regulation, the ground under the entire industry just shifted — and the days of a hands-off federal approach to U.S. AI labs are likely coming to an end.