Good evening. Here is what matters in AI today, and how to put it to work.
AI's political reckoning arrives: a Blue Wave midterm, stalling spend data, and a $1.25T OpenAI raise all land on the same day.
~4 min read · last 12 hours
In today's issue
01
Blue Wave midterms: voters reject data center expansion almost everywhere
02
Ramp data: AI spending stalls even as usage keeps climbing
03
OpenAI raises an additional $11B at a $1.25 trillion valuation
04
Chinese open-weight AI models nearly match OpenAI on hacking benchmarks
05
ChatGPT silently routes tool calls between connectors, prompting a bug bounty
Main story
Blue Wave midterms: voters reject data center expansion almost everywhere
Democrats won historic midterm victories and voters passed restrictions on data center development in 40 out of 41 ballot measures across the US.
Why it matters: If those restrictions hold and spread, they directly constrain where and how fast AI infrastructure can be built, raising costs and timelines for every team planning capacity expansion.
What to watch next: Watch whether the new Democratic majority moves quickly on data center permitting restrictions and federal AI oversight legislation, which would directly raise infrastructure costs and compliance burdens for every team building at scale.
We see three data points that together reframe the AI growth story: a landmark election result, a spending plateau, and a record-setting raise that now has to prove itself against both.
US data center ballot measures voted on Tuesday, with restrictions passing in 40 of them · TechCrunch
Watch · On the feeds
CoreWeave AI Training: Built for Training at Scale with Confidence
CoreWeave
LifeVal: Measuring AI progress on the tasks that matter in everyday life.
OpenAI
The Signal
Two forces are converging to reshape the AI investment thesis: voters just handed Democrats historic midterm wins and rejected nearly every data center expansion on the ballot, while Ramp's spending data shows AI costs plateauing even as usage climbs. At the same time, OpenAI is closing an $11B raise at a $1.25T valuation, a number that now looks harder to justify if regulatory headwinds tighten and price competition keeps compressing revenue. For engineering and product leaders, the practical question is no longer just "what can we build" but "what will it cost to operate, and under what rules."
All the best, the KYFEX team
Quick hits
AI's political and financial reckoning
Ramp data: AI spending stalls even as usage keeps climbing
Spending on AI tools has plateaued among the biggest enterprise buyers, driven by price cuts and tighter cost controls, even as actual usage continues to grow.
Why it matters: This gap between usage growth and revenue growth is the number investors and CFOs will interrogate hardest, and it should inform how you frame AI ROI internally right now.
OpenAI raises an additional $11B at a $1.25 trillion valuation
OpenAI has reportedly secured an additional $11 billion in a funding round expected to close within weeks, pushing its valuation to $1.25 trillion.
Why it matters: A valuation at that level requires sustained revenue hypergrowth, and the combination of political headwinds and spending plateaus makes that a harder story to tell to the next round of investors.
AI security risks moving from theory to active exposure
Three separate stories this week show AI-related security gaps becoming concrete, exploitable problems, not future hypotheticals, and they span models, tools, and data pipelines.
Chinese open-weight AI models nearly match OpenAI on hacking benchmarks
An experiment showed DeepSeek and GLM performing close to OpenAI's most advanced (non-public) cyber model at breaking into simulated networks.
Why it matters: Capable offensive AI is no longer confined to well-resourced Western labs, so threat models for your own systems need to account for adversaries using freely available open-weight models.
ChatGPT silently routes tool calls between connectors, prompting a bug bounty
A reported security issue shows ChatGPT passing data between connectors without explicit user awareness, with a bug bounty now open for reproducing it with an ID card.
Why it matters: Any team using ChatGPT connectors in production workflows should audit what data crosses connector boundaries and whether users are informed before that happens.
This course will teach you about large language models using libraries from the HF ecosystem
Hugging Face · Free
Put it to work
Try this today
Assess your AI infrastructure exposure to new regulatory risk
I am an engineering or infrastructure leader at a company that runs AI workloads in US data centers. Given that voters in 40 out of 41 US jurisdictions just voted to restrict new data center development, help me think through: (1) which of our current infrastructure dependencies are most exposed to permitting or capacity constraints, (2) what short-term mitigations we should consider such as multi-region or multi-cloud strategies, and (3) what questions I should be asking our cloud vendors right now about their capacity roadmaps. Be specific and practical.
Why it helps: Today's election results make this a live planning question, not a hypothetical, and running this analysis now puts you ahead of the wave rather than scrambling when a capacity constraint hits.
Before you ship it
The risk
The ChatGPT connector data-routing issue highlights a broader risk: AI tools that silently pass user or enterprise data between integrations can expose sensitive information to unintended third-party services without any visible signal to the user or the operator.
Do this
Audit every AI connector and integration in your production environment by mapping exactly which data fields travel where, then enforce explicit user consent gates and logging at each connector boundary before enabling them for sensitive workflows.
Ready to ship AI, not just read about it?
KYFEX designs and builds production AI for teams that need it working, not just demoed. Tell us what you're working on and we'll bring the engineering.
