> ## Content Index
> Fetch the complete content index at: https://www.verbalprimate.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# America is squandering a once stronger lead in AI on massive large language models - that cost too much and do too little
- URL: https://www.verbalprimate.com/america-is-squandering-a-once-stronger-lead-in-ai-on-massive-large-language-models-that-cost-too-much-and-do-too-little/
- Published: 2026-09-28T20:23:29.000Z
- Updated: 2026-09-28T20:23:29.000Z
- Description: Our focus on excessively large LLMs is squandering our lead in a hypothetical AI race. VC and PE capital are desperate to cash out of the AI bubble and are willing to say anything to thwart regulation and business scrutiny. AI doomers believe in weird things.
- Author: John Gilda
- Tags: artificial intelligence, Dave McCormick, paypalmafia, A16Z, China

## China is taking advantage of our distraction

It is ironic that the yammering SillyCon Valley venture class is beating the national security drum and their boomer politcos are clapping along in time. Boomer politicos have a difficult time with the differences between AI, LLM, NLP, ML and so on. That is OK. But the venture class *does* know. And whether they are knowingly committing financial fraud or they are experiencing AI-induced delusions, 'the why' doesn't really matter. As they fantastically message that larger context windows (size of input a model can process) and more parameters (internal learned weighting) will create artificial general intelligence or super intelligence, they have convinced the fawning media that the shambolic burning of capital, real estate, energy, and environmental resources will be worth it. The race with China meme is simply to increase urgency to reduce the chance of regulation. The concept a cold war like 'race' is flawed but convenient for messaging.

## The huge frontier models are losing the commercial race so NVIDIA hugs Hugging Face

The high cost of frontier large language models is sinking OpenAI and Anthropic and is an impediment to business adoption. We see smaller and lighter architecture opensource models are winning in the marketplace. But the hyperscalers (Google, Microsoft, AWS) and the captive [neoclouds](https://macroplane.com/blog/neocloud-stocks?ref=verbalprimate.com) and NVIDIA are all shackled to the Frontier Model model world. Anthropic is racing to IPO before bad things happen and people are beginning to question whether or not OpenAI will even be able to IPO. With Anthropic delayed IPO twice now and is scheduled for after the election at [year end](https://www.forbes.com/sites/jonmarkman/2026/09/07/anthropic-delays-ipo-to-mid-october-locks-in-15-billion-credit-line/?ref=verbalprimate.com). I doubt they will really IPO this year. Regardless. market forces are drifting away from the frontier foundries that are intent on striving for artificial general intelligence (AGI) or broad super intelligence and people are moving using lighter opensource models.

> **Core thesis:** In 24 months, the LLM landscape transformed from a closed frontier (GPT-4 dominant) to a democratized ecosystem — open-source models matching frontier quality, context windows growing 1,250×, and reasoning emerging as a first-class capability tier. The imitation cycle compressed from 12 months to 6 weeks. [LLM Evolution](https://www.llm-evolution.com/?ref=verbalprimate.com#charts-sec)

This thesis is a reason that Mr. 'we are NOT Enron' Jensen Huang is giddy about purchasing [Hugging Face](https://blogs.nvidia.com/blog/nvidia-to-acquire-hugging-face/?ref=verbalprimate.com). The trade press supplicants are lathering in Jensen's brilliance. And there is not much WTF. Here are a couple of examples:

> Against roughly $150 million of annualized revenue, $12.9 billion works out near 86 times sales and close to three times the $4.5 billion Hugging Face carried after its 2023 Series D. Nvidia joined that round, alongside Google, Amazon, Salesforce, IBM, Intel, AMD and Qualcomm. A multiple like that gets paid for a position in a market, and the position here is the point where developers choose what to run. - [Forbes](https://www.forbes.com/sites/jonmarkman/2026/08/27/nvidia-strikes-129-billion-deal-to-buy-hugging-face/?ref=verbalprimate.com)

> For Huang, Nvidia is benefiting from the growth of both parallel paths Delangue mentioned. Ultimately, the more people use A.I., the bigger the demand for its chips. “The world will need both closed models and open models. And both closed models and open models are skyrocketing in use…and [they’re simultaneously driving our sales,](https://www.fool.com/earnings/call-transcripts/2026/08/31/nvidia-nvda-q2-2027-earnings-call-transcript/?ref=verbalprimate.com)” he told analysts on Nvidia’s quarterly earnings call last week. - [The Observer](https://observer.com/2026/09/nvidia-hugging-face-acquisition-open-ai/?ref=verbalprimate.com)

86 times annualized revenue (whatever is meant by annualized revenue in this case.) That is a lot. In technical financial terms, **it is a shit ton**. Consider the S&P 500, that in 2025 averaged around 3.3 time sales. Doesn't that valuation demand a question and response on why? Or an observation about how that valuation relates to other large revenue to price ratio acquisitions. But I get it. It was labor day weekend and the financial and technical press were tired from copying and pasting press releases and announcements. Not lazy...just exhausted. Or perhaps just worshipful of the rich tech lordlings. For comparison, here are some industry average price-to-sales rations for specific segments:

![](https://storage.ghost.io/c/41/3a/413ab923-aaaa-4b3c-acf3-a8d625d98ab8/content/images/2026/09/Screenshot_2026-09-07_08-40-17.png)

Price-to-Sales Ratios by Industry 2026 - [Eqvista](https://eqvista.com/price-to-sales-ratio-by-industry/?ref=verbalprimate.com)

Lina Park over at Forkast has a view of this being a vertical integration play by Jensen.

> When the entity that controls the supply of high-end compute also owns the distribution layer for open-weight models, the incentives for gatekeeping become structural. This shift is particularly stark given that Hugging Face previously rejected a $500 million investment from NVIDIA earlier in 2026, citing concerns over the influence of a single dominant investor — a concern that has now been rendered moot by an outright acquisition. - [Forkast](https://forkast.news/nvidias-12-93b-hugging-face-acquisition-becomes-definitive/?ref=verbalprimate.com)

Sure, Hugging Face sought out the deal, according to the CEOs. Hey, we had a thought, visited the Wizard of Oz, you know behind the curtain. And badaboom, badabing bags stuffed with cash and no flying monkeys.

> 'Yes. So, I mean, during the summer I think we realized that Hugging Face and open source AI in general was at a turning point and that it needed more resources, more scale, more visibility. So, we went to see Jensen, and we told him, we want to make open source AI big. And he told us, let’s do it. So, that’s really how it started. And a few weeks later, here we are.' - [CLÉMENT DELANGUE](https://www.cnbc.com/2026/09/03/cnbc-exclusive-transcript-nvidia-founder-ceo-jensen-huang-and-hugging-face-ceo-clment-delangue-speak-with-cnbcs-becky-quick-on-squawk-box-today.html?ref=verbalprimate.com), CEO Hugging Face in CNBC interview with Jensen

### Is it catch and kill or 'embrace and expand?' And really what is the difference?

Is it just another day of big tech stifling potential competition with cash? In the almost 2010s, when I worked at a mid-sized system integrator and reseller, IDC analysts were popular at conferences. They had a hot deck that said big tech was so cash and cash flow rich that their distribution channel control and competitive landscape control would become insurmountable. IDC expressed that it was possible that big tech would just buy disruptive innovators and let them die starved and wandering in the Big Co. maze. Granted big tech is rapidly taking on debt and growth is slowing - hence the need for something, anything to keep the music from stopping.

The piles of cash they were talking about in 2008-2010 look laughingly small and the potential for triggering antitrust action seemed, sadly, much larger in those days. The endless [enshitification](https://www.youtube.com/watch?v=NCX5Yst64Hw&ref=verbalprimate.com) of tech is the outcome now. More control, less competition and pricing power that is super duper awesome. Oh, and there are not many examples of opensource and big corporate tech ownership going well. 

### The Cautionary Tale of Opensource Purchases by Big Tech

One cautionary tale is the way IBM has sucked the life out of Red Hat...with a hands-off 'merger.' Expect something similar with this transaction. This year IBM is quietly integrating Red Hat [back office](https://bizfortune.com/2025/09/business-fortune-ibm-absorb-red-hat-office?ref=verbalprimate.com). In 2018, IBM who had a long standing relationship with Red Hat, went whole hog and bought Red Hat. The CEO at the time said:

> "The acquisition of Red Hat is a game-changer. It changes everything about the cloud market...IBM will become the world’s number one hybrid cloud provider, offering companies the only open cloud solution that will unlock the full value of the cloud for their businesses.” Ginni Rometty, IBM’s chairman, president and CEO. 

The 'transforming the cloud market' idea was a pipe dream that failed. IBM is an *also ran* in the cloud business and recently investors have been nervously watching slowing growth of Red Hat software sales. In FY2019 Red Hat made $3.4BB in revenue and $512MM in operating income (GAAP) so compared to the $34BB price IBM paid for Red Hat the Price / Revenue multiple is 10X. The point is that one of the largest opensource related acquisitions in the history was at a 10X revenue multiple for a firm that had real and profitable revenue and it had a 25ish year operating record. IBM has had a reasonable run stock wise since the transaction but is faltering recently. But much of the valuation increase was portfolio adjustment and spin out of less profitable hardware and services services. Some Red Hat people who were acqui-hired told me that being absorbed by the IBM-borg was a most miserable experience, despite the hands-off promises. A clash of cultures is never fun.

![](https://storage.ghost.io/c/41/3a/413ab923-aaaa-4b3c-acf3-a8d625d98ab8/content/images/2026/09/Screenshot_2026-09-07_15-08-56.png)

Seeking Alpha News Article screen shot. Article available at [Seeking Alpha News](https://seekingalpha.com/article/4937661-ibm-big-blue-sees-red-hat-mainframe-headwinds-software-tailwinds?ref=verbalprimate.com).

That was a long way of saying that big company opensource purchases do not inevitably keep customers or turbo charge the big company. For the brave, you can also dig into the Microsoft Github transaction, which is a closer analogy in platform content (opensource developer repository.) Since MSFT took the reigns, they seem to be trying to drive its developer customers away with [stability issues](https://www.theregister.com/ai-and-ml/2026/08/17/github-has-issues-as-repo-downloads-hit-50-error-rate/5288543?ref=verbalprimate.com) and slipping [Copilot quality](https://www.nxcode.io/resources/news/github-copilot-getting-worse-2026-developers-switching?ref=verbalprimate.com). It makes me wonder how much the NVDIA | Hugging Face transaction will rhyme with these Big Co. buys opensource firm debacles.

### What are the reasons NVIDIA is buying Hugging Face?

In my personal opinion, which is based on no facts or evidence, this transaction is partially hush money for the OpenAI > Hugging Face hack. Given NVDIA has a $30B stake in OpenAI and multiple [Attorneys Generals](https://www.businessinsider.com/openai-attorney-general-preserve-hugging-face-evidence-2026-8?op=1&ref=verbalprimate.com) are investigating the hack and ordered OpenAI to preserve records, it might be convenient to have the hacked entity under control, so to speak. Additionally, it gives NVIDIA 'narrative' for smaller cheaper opensource (Chinese - often) models that are being adopted more quickly than the lumbering foundation models with the variously bolted on harnesses. The mar-com narrative is that NVDIA is building a developer community among the many developers who prefer to use something other than OpenAI and Anthropic massive models works well in the traditional Silly Con Valley 'we are just an arms dealer' motif. We don't care who buys our stack: big models, small models, 1 model, 2 model, red model, blue model - we print green on all models.

Since my first draft of this section, OpenAI seems incapable of stopping itself from hacking others. La dee da.

![](https://storage.ghost.io/c/41/3a/413ab923-aaaa-4b3c-acf3-a8d625d98ab8/content/images/2026/09/Screenshot_2026-09-28_12-52-37.png)

It is not meddling it is likely a violation of the [computer fraud and abuse ac](https://www.justice.gov/jm/jm-9-48000-computer-fraud?ref=verbalprimate.com)t. CEO and Board should be very worried. [Photo and headline](https://www.bbc.com/news/articles/cw62jje658dlo?ref=verbalprimate.com) from BBC.

### Is the Hugging Face Hack just PR Hype? It *has* gotten attention.

Does anyone find the timing of these contain the beast heroics suspect? OpenAI, falling behind Anthropic in the mar-com sphere of scary AI models doing evil things, leaves a model unattended executing looping logic that attacks Hugging Face who bravely defends itself with an NVIDIA version of a Chinese model. 

We'll get back to this but let's dwell on the super scary AI doing evil things with [Altman](https://qz.com/sam-altman-senators-openai-rogue-agent-hugging-face-073026?ref=verbalprimate.com) and [Amodei](https://techcrunch.com/2026/08/16/anthropic-ceo-says-ai-backlash-is-fundamentally-a-crisis-of-trust/?ref=verbalprimate.com) falling over themselves about how dangerous a Frankenstein they created. Come on guys. You fib all the time and cannot articulate a path to profitability and want more, more, and more capital to burn in your AI incinerator. Feels like you are hyping your upcoming and desperate IPOs on 'powfull' scary AI. 

This hype is really getting under Cal Newport's skin as he patiently defluffs the hyperbole in his '[Are We at War with AI Agent “Civilizations”](https://calnewport.com/are-we-at-war-with-ai-agent-civilizations/?ref=verbalprimate.com)?' (answer no.) Which he wrote after his first attempt to de-hyperbolic the 'hack' in '[Did OpenAI’s New Model “Go Rogue](https://calnewport.com/has-ai-gone-rogue/?ref=verbalprimate.com)”'? (answer no: it executed researcher instructions without researcher supervision and without adequate controls.) 

> 'Also in July, AI agents in an [OpenAI](https://www.therobotreport.com/tag/openai/?ref=verbalprimate.com) [cybersecurity](https://www.therobotreport.com/tag/cybersecurity?ref=verbalprimate.com) experiment coordinated [an attack](https://www.pbs.org/newshour/science/ai-agents-are-hacking-systems-without-any-input-from-humans-how-did-we-get-here?ref=verbalprimate.com) on Hugging Face. Since then, the companies have [worked together](https://openai.com/index/hugging-face-incident-and-the-road-ahead/?ref=verbalprimate.com) to strengthen safeguards, and Delangue noted that an NVIDIA version of a Chinese model helped Hugging Face address the incident.' [The Robot Report](https://www.therobotreport.com/nvidia-to-acquire-hugging-face-and-keep-ai-develpment-platform-open/?ref=verbalprimate.com)

We have been creating bastioned computing environments for decades. Remember [Stuxnet](https://www.geeksforgeeks.org/ethical-hacking/the-stuxnet-attack/?ref=verbalprimate.com)? A physical USB drive needed to be plugged in to the **air gapped network** because the Iranians didn't want their nuclear program exposed to the Internet. Because, you know, [NSA](https://en.wikipedia.org/wiki/National%5FSecurity%5FAgency?ref=verbalprimate.com). We know how to architect, engineer, and build secure environments that are not fucking attached to the public internet. The biggest 'breakouts' were the result of stealing tokens or credentials. The computer used poorly secured OpenAI Hugging Face credentials to access Hugging Face. We are also hearing that the model operated for days unsupervised. Guess the scientists or whatever we call this crack crew of 'researchers' were busy doing other things rather than watching their cybersecurity experiment. It's *almost* like OpenAI needed a marketing answer to Mythos a.k.a. 'formidable weapon.'

> 'But that same coding prowess makes Mythos a formidable offensive weapon, and Anthropic says it can outstrip all but the most skilled humans at identifying and exploiting software vulnerabilities.' - [Scientific American](https://www.scientificamerican.com/article/what-is-mythos-and-why-are-experts-worried-about-anthropics-ai-model/?ref=verbalprimate.com)

The belief system that is driving these big frontier models and these we need to make god but make sure it is a god that doesn't turn on its maker are closely tied to the rationalist and effective altruism movements. These are not science-based or engineering-based thoughts. They are just dangerous almost mystical beliefs that allow the inner circles at places like Anthropic and OpenAI to justify sucking up the worlds capital to build god machines. For those that want to understand more about these beliefs and their dispersion can read more at [Futures Garden](https://garden.johanneskleske.com/rationalists?ref=verbalprimate.com), Johannes Kleske's site which provides good primers for many of the wacko 'philosophies' haunting a portion of AI's leadership - take a look. It provides the lingo you can use to laugh at these chin-strokers. But it is not really *that* funny.

![](https://storage.ghost.io/c/41/3a/413ab923-aaaa-4b3c-acf3-a8d625d98ab8/content/images/2026/09/Screenshot_2026-09-24_11-20-22.png)

Read all about it at Futures Garden.

Johannes, in a separate note, talks how framing AI as a race is self serving to tech companies, geopolitical hardliners and deregulation advocates.

> The Space Race had a finish line. The Moon was real, and reaching it first was a definable achievement. The Arms Race had quantifiable metrics: you could count warheads. Both were contests between two superpowers with clear objectives. [Critical Futures Studies](https://garden.johanneskleske.com/critical-futures-studies?ref=verbalprimate.com) would ask: what assumptions travel with the borrowed framing? What gets imported along with the metaphor?  
>  
> What, exactly, is the finish line of the AI Race?

To me it looks like there are really two freak camps in Silly Con Valley, the cultish TESCREAL crowd and the 'we will say anything to make money' Andreessen and Sacks style pump and dumpers.

### Let's face it. The AI bubble mainly benefits the pump and dump vulture capital and pirate equity speculators and developers. And we know how this administration loves developers, bulldozers and pollution. And cheap pre-IPO equity positions.

Needless to say members of Congress are holding committee meetings, [drafting protect the AI Lords from the mean Chinese model distillers acts](https://ai-law-tracker.com/federal?ref=verbalprimate.com). And taking money from the [pro-AI PACs](https://elections.transformernews.ai/?ref=verbalprimate.com) and not just Republicans. Sure, there are two sides to current AI legislation but the press seems most interested in picking up the dramatic '[The Chinese are Coming](https://www.nbcnews.com/tech/tech-news/us-accuses-china-ai-developers-deepseek-alibaba-copying-american-ai-rcna596696?ref=verbalprimate.com)' or let's [ban super intelligence](https://futurism.com/artificial-intelligence/steve-wozniak-steve-bannon-letter-ai-superintelligence?ref=verbalprimate.com) story lines. Doesn't it seem odd that the big frontier labs, heavily financed by the smartest , biggest, and most handsome tech firms and tech capitalists didn't protect their super cool awesomely awesome super valuable intellectual capital from distillation? 

Let's put aside the track record of these huge frontier models openly stealing the world's creative output for training and their ongoing lobbying to do away with certain copyright protections for 'those people's' intellectual property. Fair use?

![](https://storage.ghost.io/c/41/3a/413ab923-aaaa-4b3c-acf3-a8d625d98ab8/content/images/2026/09/image.png)

Forbes [article](https://www.forbes.com/sites/virginieberger/2025/03/15/the-ai-copyright-battle-why-openai-and-google-are-pushing-for-fair-use/?ref=verbalprimate.com) from 2025.

Distillation is not some new magical Chinese innovation that malevolently crept up on the AI buffoons. It is a known technical domain whose roots go back two decades and tools have been *very* opensource available since 2019.

![](https://storage.ghost.io/c/41/3a/413ab923-aaaa-4b3c-acf3-a8d625d98ab8/content/images/2026/09/TMAEfVnx8zBPga48BdwZ4.png)

Timeline from [Sergio Paniego](https://huggingface.co/sergiopaniego?ref=verbalprimate.com "Sergio Paniego")'s article ["A brief history of distillation in AI'](https://huggingface.co/blog/sergiopaniego/brief-history-of-distillation-in-ai?ref=verbalprimate.com)

Understanding that this type of machine learning makes smaller cheaper models from big expensive models might have triggered some big brained AI freaks to consider how to protect their models. But the hype machine won over the protect the model machine and the models were opened up to the world. Tough problem but there are trade offs that could and should be considered when big tech is inexplicably planning on spending [$4-8 trillion dollars](https://www.goldmansachs.com/insights/articles/tracking-trillions-the-assumptions-shaping-scale-of-the-ai-build-out?ref=verbalprimate.com) on your buckets of bolts massive LLM systems. This is not a show up with you poster to the gym for the science fair to get a ribbon for participating situation. Act like adults you freaks.

And now, an Anthropic researcher flees, tweets some scary stuff and gets his 15 minutes of fame maybe as he soft launches his yapper and substack brand. And says 'I am not a doomer' while saying doomer things in a Wired interview:

> Imagine you versus a monkey. AI has the same sort of difference in intelligence to a human as we do to a monkey, which I think is quite an extreme intellect difference.  
>  
> The current plan is to solve \[alignment\] at speed in the next couple of years, probably making heavy use of automated AI safety researchers. The plan is literally to make some pretty smart models in the next year that can basically do safety research and get a whole swarm of them running in parallel. Tell them, “Go and solve the whole problem of safety,” and use them to do the safety training for the next model.

![](https://storage.ghost.io/c/41/3a/413ab923-aaaa-4b3c-acf3-a8d625d98ab8/content/images/2026/09/Screenshot_2026-09-10_08-43-52.png)

[Some](https://www.telegraph.co.uk/business/2026/09/10/ai-extinction-resignation-false-flag-conspiracy/?ref=verbalprimate.com) say the researcher is just a hype man for scary 'powful' AI. Love the Wired article because the language of the 27 year old Oxbridge graduate elite researcher trying to sound like a hip Silly Con Valley anybody is quite strained. [Wired](https://www.wired.com/story/anthropic-researcher-quits-jacob-coxon-ai-fears-humanity/?ref=verbalprimate.com)

OpenAI and Anthropic and the neoclouds and NVIDIA and the hyperscalers cannot explain the current revenue and cost and projections in simple normal business terms.

Show us your current AI revenue - not some hyped-up ill defined annualized run rate and not some big co. strategic initiative top down directed 'make the world see rainbows and Pegasus' with creative AI revenue coding - real [GAAP](https://www.investopedia.com/terms/g/gaap.asp?ref=verbalprimate.com) numbers. Also show us the real costs to operate that revenue, not classifying training and retraining (normal parts of operating AI) as R&D so that you can claim some fantasy gross margins. Ed Zitron has said again and again these 'business idiots' need to start using business terms for the products.

1. What is it that your product does today, that people are willing to pay for, and what revenue does it generate?
2. What are your costs for the products - true cost of goods sold cost including operating the model and related harnesses and retrieval-augmented generation (RAG) vector databases. And no initial training is not some out of cost of goods sold number.
3. If the product is not profitable, a plan on how it will become profitable would be normal business talk.

As Ed Zitron says (and everyone even pro-AI hype-mongers should subscribe to his newsletter and listen to his Better Offline podcasts):

> "Yet for some reason — because it’s a product, I guess? — we should as a nation, society and media ecosystem should do everything we can to assure that as much money as possible is invested in *fucking *large language models* so that they can *become something they are not.**

> There is no AGI coming. There is no conscious computer. LLMs have gotten “better,” but the “better” is not the kind of “better” that actually makes “economic sense for literally anyone involved.”..." - [Ed Zitron](https://www.wheresyoured.at/the-ai-haters-manifesto/?ref=verbalprimate.com)

### And two days after their twenty something Oxbridge Researcher flees, Anthropic is warning about the evil doers using its platform.

And if the evil 'powful' story line was not hyperbolic enough, Anthropic is once again crowing about how the evil doers are doing evil things with their evil data processing. According to [The Guardian](https://www.theguardian.com/technology/2026/sep/10/anthropic-report-details-ai-misuse?ref=verbalprimate.com):

> Criminals, state-sponsored groups, spyware vendors, scientists and propagandists have attempted to use Anthropic’s powerful artificial intelligence models to design missiles and bombs, create deadly pathogens and surveil dissidents, according to a[ threat intelligence report](https://www.anthropic.com/threat-intelligence-report-september-2026?ref=verbalprimate.com#biological-misuse-sep-26) the company published on Thursday.

> “The cases we share here aren’t typical misuse, but rather examples of the most notable and novel threat activity we’ve identified to date,” [Anthropic](https://www.theguardian.com/technology/anthropic?ref=verbalprimate.com) wrote in its 154-page report. “We’re publishing this work because we believe we have a responsibility to disclose malicious misuse of our services.”

Back to the we have had [sensitive data processing environments](https://web.mit.edu/STS.035/www/PDFs/flamm.pdf?ref=verbalprimate.com) since the beginning of computing. Guess what? We did our fucking best to keep evil doers from using them. Providing public Internet access to your data processor is just a failure. Then caterwauling about how the government needs to stop you and your idiot twin OpenAI from destroying the world is stupid, and attention seeking, and of course in no way tied to your upcoming IPO that will benefit from a pop culture belief that you actually have a product that is 'powful.'

But I understand, if you cannot talk about your product's profitable revenue and the actual things it can do, these Brother Grim dark forest fairy tales are the best you can do. You are just like [Dr. Know-all.](https://www.childstories.org/en/doctor-know-all-1757.html?ref=verbalprimate.com)

(The idiocy continues every day but some day I gotta publish this blog.) Recently in - Russian developers used [Claude to program suicide drones](https://cybernews.com/ai-news/claude-anthropic-russia-kamikaze-drones/?ref=verbalprimate.com). Congrats Anthropic your inadequately secured platform is used by an autocrat to kill its democratic adversaries. And instead of waiting for the liability suits to roll in quietly with your tail between your legs you publish a 154 page report about devious threat actors abusing your beautiful franken-child. **These dimwits need their liability shields pierced.**

### Who will make money on the USA and China AI arms race?

All this 'dangerous AI' talk dovetails nicely in the we need 'powful' AI made by the U.S. And 'powful' rich, smart antidemocratic tech lordlings to protect us from our potential Chinese overlords. And it's dangerous, so dangerous, but speed is of the essence and regulation needs to be out of the question - innovation, you know. And by the way there will be nice, very nice, opportunities for the rich politico class who kneel at the altar of tech greed. Like data center speculation for you and your private equity friends at the Club and maybe if you rub the right tech lord yarbels, like a good [chia pet](https://chia.com/products/ghost-face-chia-pet?ref=verbalprimate.com), you might even get some of that [tasty pre-ipo equity](https://finance.biggo.com/news/RP-Njp4BYH%5FypPqO-9s1?ref=verbalprimate.com), like some in the administration did with SPCX.

Let's put aside the self-dealing, which we are all frankly tired of considering. Let's focus on how the invested politicos amplify the siren call of no regulation and this is all about National Security so we that are not surpassed by China. Dave McCormick, [swampster of love](https://verbalprimate.ghost.io/ghost/?ref=verbalprimate.com#/editor/post/693ed10b9f31900001b4a500), has opined [endlessly on the risk of China and AI](https://www.politico.com/news/2025/09/16/dave-mccormick-china-ai-00565303?ref=verbalprimate.com) without commenting much on the U.S of A. [profiting](https://arstechnica.com/tech-policy/2025/12/us-taking-25-cut-of-nvidia-chip-sales-makes-no-sense-experts-say/?ref=verbalprimate.com) from sales of [advanced NVIDIA processors](https://www.scmp.com/tech/big-tech/article/3365383/nvidia-ships-first-h200s-china-forecasts-no-data-centre-computing-revenue?ref=verbalprimate.com) to China. Oh, and as he wears his worry beads down and 'aligns' his **let's make more money on war** **Pennsylvania** pitch, he is screaming for [faster defense procurement](https://www.bloomberg.com/news/videos/2026-07-15/sen-mccormick-on-defense-procurement-ai-buildout-video?ref=verbalprimate.com) which will help DoD suppliers and Silly Con Valley war mongers cash coin faster. \[Not to mention Meta, where [his wife](https://www.meta.com/about/leadership/dina-powell-mccormick/?srsltid=AU7gw4XN2Q4Rz3ourdIDEulqxL03H%5FYsHH8pfghEkqnIIg1SrUBe-BsJ&ref=verbalprimate.com) is an executive.\] Keep in mind he was a hedge fund leader who led a firm notoriously [pro-China in investments that also held Russian sovereign debt](https://www.theguardian.com/us-news/2024/oct/10/pennsylvania-senate-david-mccormick-investments?ref=verbalprimate.com). His namesake Congressman Rich McCormick is also [stumping for the data center tech lords](https://cryptobriefing.com/mccormick-warns-slowing-ai-data-center-expansion/?ref=verbalprimate.com). Unfetteredman is also [warning of the China threat](https://thehill.com/homenews/senate/5877732-fetterman-ai-data-centers/?ref=verbalprimate.com). There are too many examples of the AI hawks to include.

In the end, the deranged venture capitalist and [dark enlightenment](https://cascadeinstitute.org/dark-enlightenment/?ref=verbalprimate.com) tech lords will be lucky to get out of Dodge ahead of the pitchforks, torches and chants of eat the rich. But the rest of us will be in clean up mode for many years. And the so called tech leaders of big tech, will have some s'splaining to do if they crash world capital markets. And the current administration is chock full geopolitical hardliners and deregulation hawks, so you see them berating the AI freaks for doing the 'so scary we need to slow down' meme. And Michael Burry, who is very short on AI, summarized the situation well: ["There is nothing AI to slow down."](https://www.fool.com/investing/2026/09/16/michael-burry-just-said-7-words-dismissing-ai-afte/?ref=verbalprimate.com)

Meanwhile our president is amplifying his friend Jenson's talking point, which is very important to NVIDIA's business, that everything AI, or I guess it is super intelligence now rebranded by a boomer, is just fine.

### Meanwhile China is applying AI pragmatically in areas that may give them a lasting competitive advantage.

Earlier, I wrote a bit about how China is actually doing useful things with AI in manufacturing and robotics previously in this [rant](https://www.verbalprimate.com/ai-magedon-automation-and-all-that-jazz/). But it is worth noting that last June the World Economic Forum, playground for the global business elite, wrote that we should not be surprised - in Kaiser Kuo's article aptly titled ['Why China’s AI breakthroughs should come as no surprise.'](https://www.weforum.org/stories/2025/06/china-ai-breakthroughs-no-surprise/?ref=verbalprimate.com)

> That surprise, however, is in itself surprising. A more clear-eyed assessment of China’s structural position in the global AI ecosystem — the scale of its STEM education, the density of its computing infrastructure, the maturity of its applied research base and the ability of its governance system to align investment, policy, and talent — would have made it obvious that this level of performance was within reach.

And we are facing a very hierarchical top-down communist command and control system with our softening, old, grifting techno-freaks. Sure, they helped build China's surveillance state, and spent decades playing in that huge global market, because you know, eyeballs. They even willingly dropped trou on freedom of expression, because you know, big market. Big numbers. Shiny objects. As Kaiser writes:

> One of the less visible but profoundly consequential enablers of China’s rapid advance in generative AI is the unusually tight coordination among its public sector institutions, academic research bodies and private firms. While this kind of alignment is sometimes viewed with suspicion outside of China, particularly when it is framed in terms of state-led industrial policy or state-backed enterprise, the practical effect has been to lower barriers between research and application, to accelerate funding decisions and to unify long-term technological goals across domains.

There is much chatter about how this administration's may be aligned with Russia (China's ally) and Trump's infatuation with strong man autocrats including Xi, Putin, and Un; but this administration's meddling in academic research, industrial command and control, and communist-style economic interventions seem counterproductive to maintaining an AI technological lead. The administration's actions also seems like an abdication of the free market. And a loss in faith in the superiority our ability to innovate because we are a free society. 

The atrophied and out of ideas Silly Con Valley leaders are mainly concerned with maintaining the status quo where they pump up technology start ups, cash out when they go public or caught and killed by the other club members. They also like endless coin from surveillance marketing and advertising. They see every warm body as a meat bag with [almost $1MM of data to exploit or harvest.](https://web3.foundation/hidden-price-of-free/web3-foundation-hidden-price-of-free-white-paper.pdf?ref=verbalprimate.com) Remember all the knuckled-heads walking around chanting 'data is the new oil?' Well, kinda, your data is their oil and fuel for surveillance capitalism, or worse. But back to China, it is the very expertise that fuels two sided networks and behavioral tracking for marketing and ad-tech that fueled [China' notorious surveillance state](https://apnews.com/article/chinese-surveillance-silicon-valley-uyghurs-tech-xinjiang-8e000601dadb6aea230f18170ed54e88?ref=verbalprimate.com).

> Over the past quarter century, American tech companies to a large degree designed and built China’s surveillance state, playing a far greater role in enabling human rights abuses than previously known, an Associated Press investigation found. They sold billions of dollars of technology to the Chinese police, government and surveillance companies, [despite repeated warnings](https://www.govinfo.gov/content/pkg/CHRG-110shrg45688/html/CHRG-110shrg45688.htm?ref=verbalprimate.com) [from the U.S. Congress](https://chrissmith.house.gov/uploadedfiles/2006.02.15%5Fthe%5Finternet%5Fin%5Fchina%5F-%5Fa%5Ftool%5Ffor%5Ffreedom%5For%5Fsuppression.pdf?ref=verbalprimate.com) and [in the media](https://tsd.naomiklein.org/articles/2008/05/chinas-all-seeing-eye.html?ref=verbalprimate.com) that such tools were being used to [quash dissent](https://apnews.com/general-news-93476663b0dc4e9297f8ef5ce299d9a8?ref=verbalprimate.com), [persecute religious sects](https://www.hrw.org/report/2006/08/10/race-bottom/corporate-complicity-chinese-internet-censorship?ref=verbalprimate.com) and [target minorities](https://www.hrw.org/news/2013/03/20/china-alarming-new-surveillance-security-tibet?ref=verbalprimate.com).

### And what the hell is AI dominance and how does it help anyone?

It is very ironic that these very we'll do anything to anybody for money techno-freaks are decrying the rise of Chinese AI...while many [large companies](https://msadvisory.com/list-of-foreign-companies-operating-in-china/?ref=verbalprimate.com) continue to operate in China. It is also horribly sad that our press treats them like lovable tech wizards who deserve to be rich because they are smart and worked hard. No. They believe they are the new feudal lords and we are peasants. Peasants that can support our dear techno leaders by paying rent from Clouds to SaaS to their property for rent to peasants, they love the *little* people paying them forever.

We do not hear details about what AI dominance actually is. We hear endless releases about mathematics or game playing or recently 'hacking.' But we really don't hear much about what specifically our lovely and most beautiful AI is going to improve. Coding and cyber security are low hanging fruit because of the structured domains and available nature of the data. We see good application of narrow and specific AI working on gene folding or pharma development or chemistry but those models don't need planet killing power and pollution. So what's up?

We have the venture clowns like Andreessen, Musk, and crew peddling this nebulous abundance malarkey. The abundance chafe basically says trust us, [money won't matter](https://theconversation.com/do-tech-billionaires-dream-of-shared-abundance-we-should-probably-just-tax-them-286658?ref=verbalprimate.com), everyone will have everything in abundance; as long as you don't tax us, and don't regulate us. But try as they may to either sprinkle the 'powful' scary or the magical money won't matter propaganda, we see ample evidence that they do not care about people other than their little freakzoid cult of pump and dump. Look at this ['Machine Age Fund'](https://a16z.com/the-machine-age-fund/?ref=verbalprimate.com): all they talk about is we are going to need a shit load of kit to drive AI. So that we can:

> And as we evolve from chat to reasoning to coding and other forms of knowledge work, both the demand for work and the token intensity of the work increase by orders of magnitude. Machine intelligence is going vertical. 

Chat, we know what that is. How's it work for you when you call or type your friendly business to get help? I needed to activate a SIM on my mobile network recently after the self-service activation failed and told me to contact technical support. Of course when I call I get the chatterbot who wants to direct me in some *very* wrong directions again and again until I say unkind things that its sentiment analysis understood 'as pissed right the fuck off.' So we all know how chatterbots have helped our lives. Coding - the jury is out from where I sit. But the pull [back in tokens](https://thecoelab.com/blog/tokenmaxxing-ai-budget-reality-check?ref=verbalprimate.com) as the real cost become apparent to adopters doesn't bode well for profitability of coding services. Reasoning we've been hearing about for years with forecasts of broad use in medical diagnostics and legal analysis. If adoption was full speed ahead, we'd see it everywhere things are diagnosed and we don't. Why? Because LLMs or RLMs can be opaque. Imagine sitting before an AI judge. Imagine getting a guilty verdict and some sort of penalty and not being able to know how the decision was reached. Uncomfortable and perhaps unjust, right?

Regardless, the hype leaves out the very real possibility that more efficient architectures will not require all these resources. Information technology reliably gets smaller, gets less expensive, and uses less power per unit of output. Explain to me again why the drama-Dario and Sammy-boy are swooning about their 'powful' dangerous, evil swarms a swarming. And they are telling us that they are the only thing standing between freedom and the soul sucking Chinese AI. So just fucking stop, you are the leaders of your horse shit companies. You can stop by having an all-hands call and saying 'Stop.'

But back to what does winning or losing the AI battle for supremacy mean? I guess whomever has the most 'powful' AI by default has the most 'powful' military. My suspicious mind tells me all the war robots stuff is fueled by the compound and bunker class who have realized that the hard ex-special operators in their security details will just take over the compound or bunker, dispose of the ex-feudal lord, and keep the riches and attractive slaves. The logic problem with the robotic AI driven security dream is they will then be beholden to the nerd-maintainers of robots and code - digital special operators. Same problem, just different actors. Oh, and security robots are working out so well.

![](https://storage.ghost.io/c/41/3a/413ab923-aaaa-4b3c-acf3-a8d625d98ab8/content/images/2026/09/Screenshot_2026-09-22_11-52-12.png)

Interesting Engineering [article](https://interestingengineering.com/innovation/robot-security-guard-drowned-itself-driving-into-a-water-fountain?ref=verbalprimate.com) about security robot drowning itself.

For some winning AI is all about controlling civilian unrest in a time of deep inequalities and some notice China's fusion of military and civilian AI. The Center of Security and Emerging Technology writes in their report ['Pulling Back the Curtain on China’s Military-Civil Fusion - How the PLA Mobilizes Civilian AI for Strategic Advantage'](https://cset.georgetown.edu/publication/pulling-back-the-curtain-on-chinas-military-civil-fusion/?ref=verbalprimate.com):

> As the boundaries between civilian and defense technologies blur, the United States will face difficult trade-offs between preserving the openness necessary for innovation while mitigating national security risks. Accordingly, the United States should develop a sophisticated, evidence-based approach to safeguarding research and economic security to both impede the PLA’s acquisition and adoption of advanced technologies and ensure that benign and beneficial collaborations with Chinese entities are able to continue.

At least to a casual observer, this administration has decided not to preserve openness on many dimensions and we have a grifter in chief who has significant AI stock holdings:

> The disclosure details the sale of holdings in Amazon, Meta Platforms, and Microsoft, each valued in the $5 million to $25 million range. Concurrently, the portfolio added new positions valued between $1 million and $5 million in key semiconductor firms, including NVIDIA, Broadcom, Synopsys, Cadence Design Systems, and Texas Instruments. Significant new investments were also made in Apple, Oracle, ServiceNow, and Dell Technologies. - [Trust Finance](https://www.trustfinance.com/blog/trump-shifts-portfolio-to-ai-stocks-like-nvidia?ref=verbalprimate.com)

So there is no surprise when the grifter in chief put on his grifting promoter in chief hat on recently to defend AI. This is worse than as being pitchman for the Tesla showroom on the White House Lawn. Oh and 'whoever wins AI wins' is a [Putin](https://apnews.com/article/bb5628f2a7424a10b3e38b07f4eb90d4?ref=verbalprimate.com) quote.

![](https://storage.ghost.io/c/41/3a/413ab923-aaaa-4b3c-acf3-a8d625d98ab8/content/images/2026/09/Screenshot_2026-09-15_08-15-58.png)

From PBS Article '[Trump says the only AI guardrails the U.S. needs is him as president](https://www.pbs.org/newshour/politics/trump-says-the-only-ai-guardrails-the-u-s-needs-is-him-as-president?ref=verbalprimate.com)'

The 80 year boomer, fine representative of a generation that has done quite well for themselves and less so for the kids, goes on to say he likes Flock surveillance: '[I like them](https://decrypt.co/378171/trump-likes-flocks-ai-surveillance-cameras?ref=verbalprimate.com).' He talks about them in the vein of law enforcement. But with the apparent militarization of DHS/[ICE/CBP](https://theconversation.com/ice-not-only-looks-and-acts-like-a-paramilitary-force-it-is-one-and-that-makes-it-harder-to-curb-274580?ref=verbalprimate.com), and their use of facial recognition to track protesters and other people, this might be an example of the U.S.A. being fine with degrading civil liberties for law enforcement. It is a great pitch for authoritarians to use, most of us prefer to have reasonable law and order. However, most of us believe in various rights enshrined in the [U.S. Constitution and its amendments](https://constitutioncenter.org/the-constitution/full-text?ref=verbalprimate.com). Though I digress, it is clear that this administration likes AI for war fighting and 'law enforcement' and is fine with blurring military and civilian uses like the PRC.

Mother Jones has recently did a quick article about obviously ['Killer AI Is Already Here. We’re Using It in Iran.'](https://www.motherjones.com/politics/2026/09/killer-ai-iran-autonomous-maven-poll-ai-safety/?ref=verbalprimate.com) and we have seen enough reports to know AI has been heavily used in Israeli actions. Without dwelling too much on the specifics, if maximum civilian damage is a goal, then the war AI is doing well, otherwise am not sure anyone can make a cogent case that AI has revolutionized war. But big tech seems to be making great headway selling half-baked technologies to governments. To be clear, the technology is not completely responsible for civilian devastating targeting and killing in Gaza and the West Bank. It is Israeli policy decisions and their implementation by the Israeli intelligence and military establishment, supported by U.S. and their own arms and tech. It is also sad to think of the dubious decisions by U.S. technology companies to support war fighting knowing the limits of their tech. Target marketing and target killing are vastly different but tech companies have been repurposing marketing tech since the Alexander 'log everything' days at NSA using natural language processing (a form of AI), [relationship graphs](https://www.wired.com/story/inside-the-nsas-secret-tool-for-mapping-your-social-network/?ref=verbalprimate.com) and so on for processing signals intelligence. (To be sure they are well protected by their legions of lawyers drafting protective terms.)

Meanwhile in Iran, the Department of Defense has been crowing about using AI since our first excursion to Venezuelan and gnashing their teeth about Anthropic [illegally](https://apnews.com/article/anthropic-pentagon-lawsuit-supply-chain-risk-f15e3c30186385e73e72bee82d85b05c?ref=verbalprimate.com) declaring the company a supply chain risk. Though my personal belief is Dario was again using PR to fluff up his 'powful' AI claims. Recently our commerce secretary told us that Anthropic is [“back on the right side”](https://nypost.com/2026/09/02/business/anthropic-back-on-right-side-with-trump-admin-howard-lutnick/?ref=verbalprimate.com) though the Pentagon disagreed. As soon as the Iran 'don't call it a war' started, the Pentagon was also fluffing Grok [launching 2,000 missiles](https://futurism.com/artificial-intelligence/us-military-elon-musk-grok-bomb-iran?ref=verbalprimate.com). And of course the evil child of Thiel, Palantir, is up to it's elbows in DoD contracts. In an article about Palantir's expanding footprint in the DoD, military.com writes:

> Critics, however, have raised concerns about long-term dependence on a single private contractor for critical data systems. Centralizing large portions of military data architecture within one platform can create challenges related to security, interoperability, oversight and vendor lock-in.  
>  
> There are also broader questions about how AI-driven systems are used in targeting and intelligence analysis. While the Pentagon maintains that humans remain in the loop, the increasing reliance on automated analysis can shape how decisions are made. [military.com](https://www.military.com/feature/2026/03/22/pentagon-expands-palantirs-role-ai-contract.html?ref=verbalprimate.com)

And of course [Trump](https://www.cnbc.com/2026/05/15/trump-palantir-stock-truth-social.html?ref=verbalprimate.com) and [members of Congress](https://capitolmarkets.org/tickers/PLTR?ref=verbalprimate.com) and others in the administration hold or have held shares in Palantir. But we have become numb to the self dealing. As news comes out on how Palantir was used to process intelligence and target a girl's school in Iran it is amazing that the DoD and Congress plowed ahead to shovel more money to this firm. Palantir is saying to the press that it is not their fault that bad data was processed and that the government is accountable for the data and its quality. Other reports suggest that additional 'guardrails' were put in place after the incident.

![](https://storage.ghost.io/c/41/3a/413ab923-aaaa-4b3c-acf3-a8d625d98ab8/content/images/2026/09/Screenshot_2026-09-23_09-34-15.png)

[Bloomberg Reporting](https://www.bloomberg.com/graphics/2026-iran-school-attack/?ref=verbalprimate.com) on Palantir's involvement in killing children in a school with two Tomahawk missles.

![](https://storage.ghost.io/c/41/3a/413ab923-aaaa-4b3c-acf3-a8d625d98ab8/content/images/2026/09/Screenshot_2026-09-23_09-42-45.png)

Palantir supported the over [74K](https://abcnews.com/International/wireStory/palestinian-death-toll-gaza-war-rises-74000-war-136793209?ref=verbalprimate.com) deaths in Gaza. But they are certainly not alone. The American tech firms related to AI working with Israeli war efforts include: Amazon, Cisco, Google, Hewlett Packard Enterprise, HP, IBM, Meta, Microsoft, Palantir, and Starlink. \[source [Wikipedia](https://en.wikipedia.org/wiki/List%5Fof%5Fcompanies%5Finvolved%5Fin%5Fthe%5FGaza%5Fwar?ref=verbalprimate.com)\]

This was a long section illustrating that so far winning the AI warfare battle has not been AI's forte, at least for Gaza and Iran where the civilian casualties have been extreme. To be fair, much beyond AI plays into the fighting style. Kevin T. Baker's article in the Guardian ['AI got the blame for the Iran school bombing. The truth is far more worrying'](https://www.theguardian.com/news/2026/mar/26/ai-got-the-blame-for-the-iran-school-bombing-the-truth-is-far-more-worrying?ref=verbalprimate.com) or his similar substack article ['Kill Chain.'](https://artificialbureaucracy.substack.com/p/kill-chain) The tech doesn't kill people, people kill people party line doesn't fully hold up. Many of us agree that a world of rich techno-bros getting richer selling half-baked AI 'solutions' for warfare is not abundance or the rosy happy scenario of AI making the world better.

There are real AI uses that are and can improve the world. Today these seem to be areas with discrete rule based domains where big tech China is focused: design automation, advanced manufacturing, financial services (like the [Ant Group](https://en.wikipedia.org/wiki/Ant%5FGroup?ref=verbalprimate.com)), chemical and pharmaceutical discovery. China has goals of improving productivity overall and supporting educational initiatives. But as they work toward technology independence in semiconductors and other key areas, some of their goals are conflicting. But they are intent on remaining technologically relevant probably at all costs.

### There is no rational reason to race on AI. Stop it.

The only reason race China on AI is so the pump and dump vulture and pirate equity gang can cash out before the bubble pops. It is an easy story to put in the mouths of their politicos. It echos red scare things that also reflect Republican broader 'here come the communists' (democrats) drum beat. But is it just dumb.

There is a strategic stalemate about Taiwan between US and China. Our AI relies on foundries in Taiwan to supply processors, memory, and storage to fuel the AI 'boom.' We have limited capabilities to [close the shortfall for US-based foundry capacity](https://sourceability.com/post/foundry-limits-and-policy-gaps-are-impacting-semiconductor-capacity?ref=verbalprimate.com), despite the federal blustering. This is just ground truth.

Finally, China is doing many good things to build the capabilities to go it alone in AI. But regardless, digital innovation is almost impossible to contain. Copy and paste, you know, reverse engineering too. Sure hardware restrictions have been effective is slowing China's progress, but China is both working on their tech and [trying to circumvent restrictions.](https://apnews.com/article/artificial-intelligence-china-charges-e8f5135a71b8863c66b9c73d04cb0eb2?ref=verbalprimate.com)

The fake urgency is hyperbolic commercial greed. There is no AI arms race and to the extent there is we are squandering our once more impressive lead by continuing to believe that OpenAI and Anthropic are the answer.

Pause the data center craziness. Invest prudently in use cases that actually have promise (can operate profitably.) And sic antitrust and fraud investigators on the pump and dumpers before there is a smoking hole in the ground.

> Law enforcers already have authority to charge companies and their CEOs for creating and releasing dangerous, unvetted, or defective products. We shouldn’t let discussions about new legal regimes distract from the fact that there’s no AI exemption from laws already on the books — a point [@FTC](https://x.com/FTC?ref=verbalprimate.com) emphasized repeatedly during my tenure.   
>  
> 1\. There is an extensive set of laws that govern dangerous and defective products. For example, releasing unvetted AI models or agents can violate consumer protection laws. Shipping flawed AI tools without implementing adequate measures to detect and stop rogue or defective AI agents can be an “unfair or deceptive” act or practice under the FTC Act (and analogous state laws). And some state AGs are already exploring holding AI firms and their CEOs criminally liable when their models participate in criminal activity.   
>  
> 2\. Existing laws also prohibit “unfair methods of competition.” This covers instances where AI firms appropriate the competitively sensitive information of their customers, including through tracking their use of various tools. It can also cover instances where firms pursue dangerous behavior, aware that doing so may compel rivals to do the same. As the Supreme Court has noted: “A method of competition which casts upon one's competitors the burden of the loss of business unless they will descend to a practice which they are under a powerful moral compulsion not to adopt, even though it is not criminal, was thought to involve the kind of unfairness at which the \[unfair methods of competition\] statute was aimed."   
>  
> 3\. The highly concentrated and interconnected structure of these markets could be creating major risks and conflicts of interest. We had started investigating these partnerships and cross-investments across the stack (and released a preliminarily overview of some findings: [ftc.gov/news-events/ne…](https://www.ftc.gov/news-events/news/press-releases/2025/01/ftc-issues-staff-report-ai-partnerships-investments-study?ref=verbalprimate.com)). Both federal and state enforcers should be scrutinizing these opaque relationships and inter-dependencies. We are already seeing how these relationships could undermine accountability. For example, OpenAI could face liability given the Hugging Face incident, but Hugging Face being bought up by Nvidia means that we’re unlikely to see it file a lawsuit over this — given Nvidia’s strong incentive to see OpenAI continue full speed ahead.   
>  
> 4\. As AI tools dramatically change the landscape of cybersecurity risks and hacks, all businesses should be doubling down on having core security protections in place. Firms that fail to invest in adequate data security measures or fix known vulnerabilities can also be breaking the law. A recent analysis showed that around 1/3 of Fortune 100 companies do not even have a way to notify them about security issues. During my [@FTC](https://x.com/FTC?ref=verbalprimate.com) tenure, we sued firms for poor data security practices and held CEOs liable when they were personally responsible. [this.weekinsecurity.com/dozens-of-amer…](https://this.weekinsecurity.com/dozens-of-americas-largest-companies-have-no-simple-way-to-report-security-flaws/?ref=verbalprimate.com) [ftc.gov/news-events/ne…](https://www.ftc.gov/news-events/news/press-releases/2022/10/ftc-takes-action-against-drizly-its-ceo-james-cory-rellas-security-failures-exposed-data-25-million?ref=verbalprimate.com)   
>  
> 5\. As policymakers consider new legal regimes, we should be looking to lessons from prior efforts to govern major sectors, such as banking and other networks, platforms, and utilities. Tools like structural separations, nondiscrimination, and supervision could be key, and there’s a rich history of what works and what doesn’t. But we can and must pursue any new efforts alongside enforcing existing laws. [Lina Khan](https://x.com/linamkhan/status/2099204390548639960?ref=verbalprimate.com)