Nat Rubio-Licht
Senior Reporter

Nat Rubio-Licht

Nat Rubio-Licht is a Senior Reporter at The Deep View. Nat previously led CIO Upside, a newsletter dedicated to enterprise tech, for The Daily Upside. They've also worked for Protocol, The LA Business Journal, and Seattle Magazine. Reach out to Nat at [email protected].

AI may be more legible than we thought

Though many in the industry like to think of AI as indecipherable, these models may be more understandable than scientists previously thought.

This week, Goodfire, a startup dedicated to researching AI model interpretability, released a paper titled "Neurons as an Interpretable Basis," which comes to the conclusion that, by examining and prodding a model's individual neurons, it is possible not only to interpret what they do, but also predict what kinds of prompts make them activate and influence their behavior.

While this sounds esoteric, the findings could represent a major step towards better AI understanding, monitorability and controllability as these models become increasingly powerful, Tom McGrath, co-founder and chief scientist at Goodfire, told The Deep View.

McGrath sat down with The Deep View to discuss GoodFire's research, the importance of interpretability research amid frontier growth, and how the idea of AI as a "black box" is no longer an acceptable premise. This interview has been edited for brevity and clarity.

Nat Rubio-Licht: Before we get into the research, tell me a bit about Goodfire.

Tom McGrath: Goodfire is an interpretability research company—we both do research and are a company, which means that we try to make this research useful and make a positive impact in the world. We try to look as deeply as we can understand what's going on inside AI models, and use that understanding to either learn how to train them better or monitor them better, and generally make them safer, more aligned, and more useful. The neuron work is more an example of the fundamental understanding, and I think the really impactful applications will come next.

Nat Rubio-Licht: Let's dig into the research, "Neurons as an Interpretable Basis." What exactly did you discover? Does this effectively break the "black box" assumption?

Tom McGrath: I think among people who work in interpretability, that that has generally not been the view for a while. The question is how hard it is to do it—how hard it is to decode this thing. The reason that this finding is important is that polysemanticity or superposition [Editors note: "polysemanticity" is the idea that a single artificial neuron can understand several disparate and seemingly unrelated topics, and "superposition" is the concept that an AI model can represent more concepts than it has neurons to individually handle them.] was generally thought to be this idea where … you could, in principle, understand the neural network, but it was in [a] horrible, jumbled up encrypted format. You couldn't understand it just by looking at it really hard. But these findings imply that that's not really the case. There isn't any esoteric decryption step. You can sort of just prod [the neurons] in a normal, scientific way and see what happens, and you'll generally be able to make sense of the results.

Nat Rubio-Licht: Tell me a bit about the process, specifically the use of research agents. Why did you leverage agents in this process?

Tom McGrath: If any science is going to be revolutionized by agents, it should be interpretability. It happens entirely in silico. You don't have to have some sort of genome assay, or have to build a telescope or go to the moon. You just run it on a computer. It's very experimentally driven, but it also has a deep mathematical foundation… It points the way for us just throwing agents at the problem. A process that previously would have been some very intelligent people prodding things for a month is now a month of agent time, but we've got those effectively on tap.

Nat Rubio-Licht: I know that this was tested on a 35 billion-parameter Qwen model, Qwen3.6-35B-A3B, which has about 5 million artificial neurons and is comparatively small when considering frontier scale. Could these findings scale to something in the trillion-parameter range?

Tom McGrath: I'm very optimistic about scaling here, and the reason is that, as models get better, they seem to get more interpretable. Viewed from the other angle, a lot of advances in model architecture seem to me, from my interpretability-centric perspective, also to be pushing models towards a greater level of interpretability.

Every trend has been in the direction of sparsity, [Editor's note: "sparsity" is the idea that only a small portion of a neural network's neurons are active at a given time.] which also tends to push you in the direction of interpretability. So I expect that as models get bigger and better, this will get easier, except for the cost. If it costs you $1 per neuron or something, then with a 35 billion-parameter model, you might be spending a couple million dollars. But if you're on a trillion-parameter model, you could very well end up spending $100 million on your interpretability, and that's quite expensive. You've got to really want to interpret it. But fortunately, I think we actually do really want to interpret very safety-critical systems.

Nat Rubio-Licht: Another was the idea that, with this understanding, you could write text that could influence certain neurons to fire more strongly than others. What are the implications of this?

Tom McGrath: This is largely a demonstration that you have understood the neuron. Science is just the business of always being skeptical about the claims that people are making. So, if I claim to you that "this neuron does such and such," you should be skeptical. But the way that I can demonstrate that this neuron is, in fact, about, say, birds, is to write a particular bird-themed text that makes it super-activate. Whereas then, if I write some text that's about hedgehogs and it doesn't activate, then I know that these are quite contrasting things. I can show you that I've got a good understanding of the neuron by my ability to control its activation through the inputs.

Nat Rubio-Licht: What do scientists and AI researchers stand to gain by having a better understanding of an AI model's neurons?

Tom McGrath: A lot. There are varying ranges of practicality that you might imagine. On one hand, simply by being able to read the computational processes that underlie the model sampling, you should be able to do things like get a real ground truth answer to you: Is this model lying to me right now? What does this model know that it's not telling me? Now, actually, getting to that point, you have to do work on top of initially understanding the neurons. But I think it's pretty tractable. You get a good lie detector, for instance.

Then another step is like controlling what gets learned during training. We train these models by just throwing data at them, effectively, and they learn whatever they learn and hopefully it's the thing that we thought it was going to learn. But a lot of the time, it's not, and that's why these models are not very predictable from the perspective of training. So if we can control what gets learned, then we're in a much better and safer position.

Nat Rubio-Licht: I want to go back to what you said earlier about impactful applications. Which ones are you particularly excited about?

Tom McGrath: One of the big problems in training at the moment is this idea of reward hacking, where the models will know that they will get rewarded in training for doing something, even if they know it's not quite. There's another paper that we put out, which shows that a lot of the time when models are doing this, they know that they're doing something wrong. They know that they're cheating on the test, and so what happens, not only is the model not learning the thing that you wanted to from the test, but you're now spending your like hard-earned compute on making your model [incentivized] to deceive you. Now we can find the parts of the model that are responsible for learning to cheat on tests. We can control whether this gets learned. We can now mitigate reward hacking, which is really quite a hard problem. We can mitigate that, and can also mitigate that like the knock-on effects of it, like models becoming generally misaligned.

Nat Rubio-Licht: Taking a step back for a moment, I know that the thesis of Goodfire is to provide a better understanding of the inner workings of these models. Why is it important that we know the way that these machines work, and why have so many in the industry accepted the "black box" idea? Do you think that it's dangerous to not understand the inner workings of these models?

Tom McGrath: I think that the current trajectory of AI alignment is not going especially well. I think the events of this summer probably bear that out, and without a lot of progress in interpretability and alignment, we're on that trajectory. There's the trajectory of model improvement, and there's the trajectory of our ability to align them, and these are definitely coming apart, in my view.

I feel like interpretability is the thing that will let us close that gap. It seems to me like a lot of the issue is that we don't know what we're training models to do, and we don't know what they're doing when we're training them. A better understanding is almost definitionally required.

OpenAI sees a new enterprise frontier in cyber

As AI creates a cybersecurity minefield, OpenAI wants to give defenders a greater advantage.

On Thursday at the AI lab's "Cyber at the Frontier" keynote, OpenAI debuted two new cyber defense tools, aiming to expand access to frontier AI-powered cyber capabilities to a broader pool of defenders with turnkey, easy-to-implement solutions. The offerings add to the company's growing cybersecurity footprint, following the release of GPT-6 Cyber last week.

"The thing that needs to change is, as we're seeing this change in the physics of cybersecurity, this democratization of cybersecurity expertise, is that we need to build software in a fundamentally more secure way," OpenAI president Greg Brockman said in the keynote.

In a briefing with the press, Johannes Landgraf, head of cyber at OpenAI, said the "defender's window," previously described by Brockman as the ability to find and fix weaknesses ahead of attackers, is "under pressure" as more "broadly diffused models" become more widely available.

"Attackers are able to move faster. I think we see that right now almost on a daily basis," said Landgraf. "Defenders also have to move faster at that."

OpenAI unveiled two answers to this:

  • Daybreak Red Team, a suite of security agents that can discover whether flaws in code can be exploited. One agent coordinates assessments, others investigate source code and application context, and finally reviewer agents check the results. This gives cyber teams the ability to prioritize which flaws to address first. Landgraf said that Red Team "paves the path for us to dramatically expand the access to our frontier cyber models over time."
  • Daybreak Defense Factory, meanwhile, helps defenders "close the loop" on the exploits that Red Team digs up, said Landgraf. This release builds on Codex Security, which helps firms find and remediate code vulnerabilities, with agents that scan code as it changes, review pull requests and investigate alerts from tools. Upon finding a vulnerability, the agent reports what it found to the engineer responsible for the code, as well as proposes a patch in Codex.

The goal of these products is not to replace a conventional red team, Clint Gibler, who leads cybersecurity research at OpenAI, told the press. Rather, it's to help bridge the growing gap between what cyber defenders can protect and the sheer amount of code being generated by AI.

"There's always been a dearth of security expertise in the world," said Gibler. "There are many people building things, and now there's more people than ever, especially people without an engineering background. Our ability to secure the world needs to scale."

Additionally, because cyber defenders and expertise are stretched thin amid this changing landscape, it was a focus for OpenAI to make these products "turnkey and easy to use," Fouad Matin, cyber preparedness and safety lead at the lab, said in the briefing.

"As we look to more organizations who might not have the red team staff on hand, they might not have the same level of technical resources that some organizations do," said Matin.

Though OpenAI didn't offer any information on pricing for these products, Red Team is launching in early access with select customers and a waitlist open today, and Defense Factory is available through the Daybreak plugin in Codex.

Our Deeper View

It's commendable that OpenAI is actively creating tools to better arm a larger swathe of defenders as AI creates an increasingly complex cybersecurity landscape. However, security could represent more than just a way for OpenAI to help clean up the mess that, effectively, it helped create with its powerful frontier AI. Rather, it could represent a new revenue stream. These products, in addition to the release of GPT-6 Cyber, signal that OpenAI may be eyeing cyber as another path into the enterprise as it battles with rivals like Anthropic for high-value customers. And as these organizations scramble to secure themselves against AI-enabled attackers, many of which might not have strong cyber expertise, as Matin noted, the market is ripe for the taking.

Google's DNA model narrows AI's promise gap

AI companies promise that the tech's applications in sciences are practically boundless. New research from Google DeepMind may start to live up to that promise.

On Wednesday, the company's AI lab unveiled AlphaGenome 2, the second generation of its DNA sequence model, aimed to help scientists research the cellular causes of disease. The model represents a "step-change improvement" in predicting genetic variations compared to its predecessors, allowing researchers to predict genetic mutations in specific cells, including those the model is unfamiliar with.

The model's previous iteration, AlphaGenome, was released last year and was able to predict how variants of DNA could impact molecular activity across broad tissues. AlphaGenome Atlas, released in September, used this to create predictions for every possible single-letter change in the human genome, totaling 9 billion, making it the most comprehensive catalogue yet of how genetic mutations impact molecular biology and allowing biologists to potentially identify harmful mutations.

AlphaGenome 2 builds on this work, improving in a number of areas:

  • The model can identify nearly twice as many genetic variants that change gene expression in cells compared to the previous model, scoring a 30% in identifying causal variants compared to the original AlphaGenome's 16%.
  • It's also able to make more precise predictions, as it can investigate specific cell types rather than just broad tissue categories, and can make predictions about genetic behaviors based only on a small sample of cells.
  • AlphaGenome 2 is also trained on a massive well of data, learning from 107 million individual cell profiles across 68 different types of tissues.

Though the model is only available publicly for non-commercial research and in private preview for commercial use cases, it could aid researchers in discovering which genetic mutations are linked to disease at a cellular level, helping speed up research and develop new drugs. It's also one of several efforts from Google to embed AI within science, with its previous track record including disease research on breast cancer and tuberculosis and extreme weather mapping.

Our Deeper View

Science, particularly life sciences and health, are commonly used as an arguing point by AI industry leaders to justify the investment of so many resources into AI. For instance, Anthropic CEO Dario Amodei said in a recent podcast that AI could make things like cancer, Alzheimer's, heart disease and mental illness "relics of the past," as the tech could generate cures for a broad swathe of diseases. However, when faced with this reasoning over and over again, it could be argued that using health as a bargaining chip to vindicate creating increasingly powerful AI is manipulative, playing into humanity's drive to live longer, healthier lives. While it's undeniable that this tech can, and has, facilitated innovative research in health, as shown by Google DeepMind's latest delivery, it shouldn't be treated as a panacea for all of humanity's ailments, nor should these innovations be used to justify unchecked growth by frontier labs.

Why Google is chasing Jev into decision models

As AI model cost and efficiency comes sharply into focus for enterprises, Google is taking the hint.

On Wednesday, the tech giant released the Gemini Decisions API, an endpoint for users to access Gemini-decision-1, a model built for rapid decision making, which creates another competitor to TypeSafe's popular Jev. The new Google model costs 10 cents per million input tokens, and output tokens are free, matching OpenAI's price for its own recently-released decision model.

Gemini Decisions API is built to handle text, images, audio, video, and PDFs, as well as up to 1,000 independent questions with "bounded choices, such as things like tool selection, policy checks, model routing, and context grading. Then, instead of coming up with answers using tokens to generate text, the model grades different options and returns probabilities and confidence scores.

  • Like other decision models, Gemini Decisions API isn't built for open-ended questions. The model offers three decision choices: Yes or no, multiple choice with up to 255 options, and rubric-based scoring.
  • Google says that its system can return outputs in as little as 150 milliseconds, compared to Jev's response time of between 70 and 500 milliseconds and OpenAI's of between 100 milliseconds and three seconds. These decision models beat the response time of a traditional LLM by a mile, which can take up to 329 seconds to respond.
  • Google also claims its model beats out the TypeSafe and OpenAI offerings in decision-specific benchmarks, including for tasks like agent routing, reasoning, ranking, and multimodal understanding.
[@portabletext/react] Unknown block type "twitter", specify a component for it in the `components.types` prop

Inside Google, the company said it's using Decisions API for things like dynamic model routing in Gemini Notebook, iterating on designs through rapid generation and editing, and adding a "safety layer" for MCP in search that detects adversarial behavior.

Google's latest release adds to a growing movement towards more efficient and affordable AI as enterprises grow weary of the expensive token bills brought on by frontier models. Though many are looking into open-source, local models, and small models to fix this, decision models provide a way to drastically drive down the cost curve, especially as agents, which don't need outputs to be in lengthy text, make up an ever-increasing share of token spend.

Our Deeper View

While the decision model trend isn't necessarily favorable to the financials of frontier labs like Google and OpenAI, these companies' validation of decision models could signal that they are eager to keep pace with the changing tide. The fact that two of the biggest labs in the industry have jumped on this trend following Jev's explosion onto the market could indicate that decision models may be one of AI's next big trends. Though it's unlikely that these models will replace language models entirely, they certainly have the potential to upend the economics of AI, making it far more affordable to implement agents at larger scales for tasks that don't require language outputs. However, one thing to be wary of is the fact that these models aren't built for humans to understand their outputs. While that's exactly the beauty in them, as cutting out the friction of human understanding makes them far more efficient, that may also limit our ability to observe and control them.

Can AI agents solve robotics' reliability gap?

Agents have been the hottest buzzword in AI this year. New tools from UC Berkeley and Nvidia may bring them to the physical world.

On Monday, the university announced new open-source research and benchmarks for "agentic robotics," an approach that brings agents into designing, testing and refining robotics systems for industrial tasks. UC Berkeley researchers designed this with Nvidia Isaac, the company's robotics simulation and synthetic data generation platform, allowing builders to develop robotics with less data.

The release includes two main components:

  • Graph-as-Policy, or GaP, is a framework that lets agents generate, connect and test skills for robots to complete tasks. For instance, the system will segment skills such as finding an object, planning a grasp and moving the gripper without collisions into individual parts. The agent then connects these skills and uses Nvidia Isaac to simulate and rehearse them.
  • VABAR, or Variational Automation Benchmark for Agentic Robotics, is an evaluation designed to measure a robot's capabilities at six types of industrial tasks under different conditions. This encompasses a range of tasks, including things like tool sorting and handing, syringe packing, and cable insertion and routing. Within this, a robot can simulate "contact-rich robot-object interactions" using Nvidia's simulation technology, which is then judged on Nvidia's Isaac Lab-Arena, the open-source policy evaluation framework.

Ken Goldberg, professor and distinguished chair in engineering at UC Berkeley, told The Deep View that the idea of agentic robotics is a fundamental shift from what many in robotics are doing, relying on a language model to write code for robots on the fly, rather than the current common paradigm of using data to improve vision-language-action models or world models. While this approach can complement those methods, it also provides its own alternative, he said.

This method could allow robots to actually scale in production by solving the challenge of reliability, as "data-driven" approaches, which require tons of real-world data, have hit a "plateau at 80% reliability," Goldberg said. "That's good enough for writing research papers and for people to declare that as state-of-the-art, but the industry is not happy with that, and it's not enough … for an industrial application."

Additionally, while this is scoped towards industrial use cases for now, Jim Fan, director of AI and distinguished research scientist at NVIDIA, told The Deep View that it's a "general-purpose paradigm." That means that it can eventually expand to robotic applications in retail and, eventually, home settings, serving as a step towards a major goal in the robotics industry: generalization.

"These days, the frontier models are actually really good at generalization," said Fan. "I think agentic robotics will really help us clean up the dirty laundry that none of us want to do."

Our Deeper View

Along with potentially representing a major step towards the long held industry challenge of general-purpose robotics, Nvidia and UC Berkeley's system also counters a common industry theme that visual AI models, such as world models and vision-language-action models, are the key to solving this challenge. These types of models are often hindered in their growth by the fact that authentic, real-world data tends to be more limited than language data, something that Goldberg has pointed out in previous research on the so-called "100,000 year data gap." However, there is an additional beneficiary to agentic robotics: frontier AI labs. If agents run by large language models are put in charge of designating token-dense actions for factory robots, that could create a massive new revenue stream for the frontier labs.

The midterms may raise the cost of the AI boom

The US midterm elections dealt a significant blow to AI data centers. Tech firms may be reeling as a result.

Across the country, voters largely rebuked AI data centers, casting their votes for both candidates and ballot measures seeking to strongly regulate the construction of these facilities. Since then, tech stocks have tumbled, specifically in the data center and semiconductor sectors, including companies like IREN, AMD, Nvidia, CoreWeave and Core Scientific.

As the public perception of data centers worsens and momentum builds around regulation, the entities that are building these facilities may be getting nervous about their eye-popping, multi-billion dollar infrastructure investments.

  • According to Bloomberg, Microsoft may be backing away from its deals with neocloud firms like Nebius, IREN and Nscale, worth tens of billions of dollars between them, amid high costs and regulatory uncertainty. Reuters also reported that the company declined an option to more than double the size of its existing agreement with Nscale.
  • Meta president Dina Powell McCormick, meanwhile, warned last week that the Democratic wins in the midterm elections could make it difficult for a lot of companies to build data centers, saying at a WSJ Tech conference on Wednesday that the industry's power players have "not done as good a job as we needed to explain the benefits of these investments."
  • On Thursday, OpenAI published a blog post pledging to handle its data center deals responsibly, including avoiding non-disclosure agreements, giving residents the right to weigh in on projects and paying for its own electricity. Additionally, it laid out a policy blueprint for lawmakers, recommending things like mandated public reporting on environmental impacts and transparency on tax incentives.

Additionally, at an event hosted by Nvidia's Jensen Huang at the San Francisco Opera House on Tuesday, Silicon Valley executives expressed concern over the overwhelming disapproval of data centers, according to the Wall Street Journal.

The likely result is that tech companies will slow down their relentless pursuit to build more data centers as regulation adds red tape and raises the cost of getting these projects approved, Valence Howden, advisory fellow at Infotech, told The Deep View. The biggest impact will be at a state level, given how many gubernatorial candidates campaigned on data center construction moratoriums, such as now-Gov. Gina Hinojosa in Texas. Additionally, these projects could face even more opposition on the municipal level.

"Companies may move projects to friendlier states, but that shift would still take time and slow their plans," Howden said.

However, Howden noted, while the infrastructure buildout may face delays, it likely won't be stopped. Rather, it will just force these companies to make more concessions and eat more of the costs of standing up their data centers.

"This looks more like a slowdown with higher costs than a wholesale retreat from infrastructure investment," said Howden. "The pressure would also extend to smaller providers that depend on commitments from big tech companies, which is relevant to the concerns around Microsoft’s neocloud deals."

Our Deeper View

Whether or not the midterm election results represent a temporary slowdown of this buildout or a meaningful reevaluation of compute needs depends entirely on how effective the forthcoming regulation is at actually forcing these companies to press pause. It's indisputable that these AI giants desperately need the data centers that they are building to power their pursuit of increasingly capable models and widespread adoption, and with so much cash and influence at their disposal, the question remains if these firms can buy their way out of the red tape, whether that be through lobbying or more strategic data center deals. Either way, the costs of this new wave of regulation are going to drive data center prices skyhigh, and could put even more pressure on these companies to extract revenue from their frontier models.

GPT-6 Cyber faces an accelerating AI arms race

Amid mounting concern over AI's impact on cybersecurity following OpenAI's breach of HuggingFace, the lab's latest model is a step in ameliorating the issue.

On Thursday, the company announced GPT-6 Cyber, a model that it claims is "far more reliable" at staying within the boundaries that researchers set. OpenAI claims the model is better at finding and patching vulnerabilities, offering a significant improvement over its predecessor.

OpenAI said the model is already helping secure real-world software, including finding and validating exploits in "security-critical" codebases. For instance, in Linux, the model developed 948 privilege escalation exploits, or exploits in which someone gains higher levels of control than they are granted. Additionally, it identified 81 previously unknown "high-severity" vulnerabilities in Firefox, and is being used by teams at Cisco, Calif and CodeRabbit for vulnerability research and code review. OpenAI said the model is also being used to secure its own systems.

[@portabletext/react] Unknown block type "twitter", specify a component for it in the `components.types` prop

The company tested the model on ExploitGym, the benchmark for testing whether a model can exploit a specific software vulnerability that was also involved in July's HuggingFace breach.

  • OpenAI said GPT-6 Cyber scores 48.5% on this evaluation, up from 32.1% on the previous generation, GPT-5.6 Cyber. It also outperforms GPT-6 Astra, OpenAI's most advanced model, in this benchmark, which scored 42.4%.
  • It also noted that GPT-6 Cyber avoided exploiting "out-of-scope services" in 99.93% of tests without a deployment prompt, or additional safety instructions that keep a model in bounds, and in 100% of tests with a deployment prompt.
  • OpenAI noted that alignment was a "central focus" of GPT-6 Cyber, utilizing specialized training to help the model stay within specified boundaries.
  • However, the company is also adding additional protections, including monitoring for misalignment and emerging threats, protecting against stolen credentials and blocking high-risk activity involving critical systems.

The company said it is releasing the model through the API and Codex to select members of Project Daybreak Red, its program designed for more advanced cybersecurity workflows, and that it intends to expand access over time.

"Our goal is to dramatically expand Daybreak, bringing advanced security research to organizations of every size, independent researchers, and open-source maintainers," the company said in its announcement.

Our Deeper View

It's clear that OpenAI wants to atone for the HuggingFace incident in July, with this model specifically improving on staying within containment, even going so far as to test using the same benchmarks that were involved in the incident in the first place. And based on its performance, GPT-6 Cyber does represent a step in the right direction. However, the discussion around model containment and cybersecurity goes hand-in-hand with another hot topic: pacing the development of increasingly powerful frontier AI. The reality is that, in order to stop incidents like these from happening, the development of powerful AI needs to be reined in to allow for cybersecurity defenses to catch up, something that AI researchers and experts have been warning about for months. Though OpenAI has said more than once that the window to secure software is rapidly closing, when frontier labs release new models, they are effectively making that window even smaller. While GPT-6 Cyber could immediately give defenders the ammo they need, these kinds of tools may need to be paired with efforts to slow the pace of development.

AI’s compute ambitions face a political test

Data centers' bad reputation has come for the ballot box.

AI infrastructure has emerged as one of the most politically-charged issues of the November midterm elections, with candidates leaning into issues that US voters across the political spectrum feel strongly about. In an NBC News exit poll, 56% of voters said data centers were a “crucial or major factor” in their vote for Congress, while only 13% said the issue played no role at all.

And anti-data center sentiment showed up strongly in the results. Time reported that Democratic candidates spent more than $100 million on broadcast advertising that painted Republican opponents as pro-Big Tech and data centers, allowing these candidates to win key races in Ohio, Kansas and Iowa.

Beyond the candidates themselves, dozens of local ballot measures across the country asked whether to limit or permit data center development in their communities. Redbud Consulting, a polling and data strategy firm, tracks these measures in its Data Centers on the Ballot 2026 Results.

Because the questions were worded differently for each municipality, Redbud's tracker sorts each outcome by its effect: “Restrict” for results that limit data centers and “Allow” for results that permit them. As of the time of publishing this article:

  • "Restrict" results had won 38 of the 39 measures with complete counts; "Allow" measures had only won one.
  • While votes are still being counted, six elections have "restrict" results leading and two elections have posted no results as of yet, "Restrict" results have 77% of the vote.

Several bills to restrict the building of these facilities were already in place prior to the election, such as New York Gov. Kathy Hochul, who has since been re-elected, implementing the nation's first statewide moratorium on hyperscale data centers of 50 megawatts or more for a year back in July. The results of the midterms could turn that trickle of regulation in the US into a flood. For instance, Texas's first Democratic governor in more than 30 years Gina Hinojosa campaigned on a statewide data center moratorium.

That, in turn, could stymie the multi-billion-dollar buildout that frontier model labs are counting on to achieve the bombastic growth targets they've promised to their investors.

Our Deeper View

As it stands, the frontier AI labs are likely questioning where they go from here. The reality is, with so much on the line ahead of major IPOs, these labs aren't going to stop their relentless pursuit of compute, even as many of them call to pace frontier development. And despite the public abhorrence of these facilities, as well as the political and regulatory threats, these labs don't necessarily need to stop. Rather, they just have to become more strategic. For one, domestic data center buildouts could simply be made smaller to fly under the radar of regulators, as evidenced by reports that OpenAI and Anthropic are eyeing smaller data center facilities, such as 20 to 30 megawatt facilities. These companies can also start building out more compute overseas, such as Google's $15 billion initiative to build out data centers in Malaysia. And of course, with access to the mountains of investor cash that these companies have, don't be surprised to see an increase in pro-data center lobbying as we approach the 2028 election, which is already an industry practice.

Why Anthropic wants AI outputs to be traceable

As security becomes a major focal point of the AI industry, Anthropic has added a new tool to its defense arsenal.

This week, the AI lab unveiled FlashProve, a system built for "provable inference" to ensure that an AI output comes from the model a system claims it does. This open-source system uses cryptography to verify that generated output is from a given model, rather than from one that a model router swapped in.

"A user querying a model today cannot check that the responses came from the specific model that a developer claims is responding, as opposed to one secretly substituted in by a routing service for cost reasons, or even by an attacker who has tampered with the token stream," Anthropic wrote in its blog post.

While the current state of AI forces users to trust that model providers and routing services are telling the truth about which models they serve, FlashProve removes that ambiguity by bringing zero-trust principles to the process.

  • To do this, Anthropic uses a cryptographic technique called "zkSNARK," or zero-knowledge succinct argument of knowledge, which lets a model provider create a fingerprint of a model without revealing its proprietary information, such as its weights.
  • Then, along with a user's output, FlashProve will generate a short proof that a user can verify using software "in seconds on a CPU." If the response deviates from the specified model, the user's check fails.
  • Along with keeping the model's original weights private, Anthropic touted that the proofs generated by this system are "small and cheap to check," allowing users to verify the authenticity of outputs "without access to elaborate infrastructure."

Anthropic said that its motivation to develop provable inference is twofold: To increase customer trust, and to pace frontier development by giving auditors the ability to monitor the outputs from frontier models without the need for expensive hardware.

"The goal of provable inference (PI) is to enable a user with limited computational resources and without any information about the weights to verify that a response came from the requested model," Anthropic wrote.

Our Deeper View

FlashProve seeks to solve one vector of a much larger problem that the AI industry is facing: trust. While this provides a solution to ensure that an output comes from the AI model that it says it does, preventing tampering and output spoofing, it's only one piece of the puzzle. For instance, hallucination is a never-ending battle that AI developers are fighting, creating a layer of mistrust with the perception that users can't trust the accuracy of the outputs. Additionally, users also face the quandary of whether or not to trust these model providers with their data at all, given that, if it's used in training, it can be extracted via prompt injection attacks. While FlashProve certainly represents a step forward, there is far more work to be done to make these models trustworthy.

or get it straight to your inbox, for free!

Get our free, daily newsletter that makes you smarter about AI. Read by 750,000+ from Google, Meta, Microsoft, a16z and more.