Anthropic's AI gatekeeping, regulatory capture fears, California election integrity, and inflation warning signs
Anthropic released a new flagship model (referred to as "Fable 5") that tops benchmarks but comes with significant policy strings attached: mandatory 30-day data retention even for enterprise zero-retention customers, secret model downgrades for users Anthropic deems unworthy (only partially walked back after public pressure), and new restrictions on genomic research and frontier AI development use cases. The speaker argues these moves collectively reveal a regulatory capture agenda—Anthropic lobbying for model-level regulation that would crush open-source competitors while consolidating its own dominance.
A central concern is the practical fallout: life-sciences and genomics companies being forced off Anthropic's APIs are turning to the best available open-source alternatives—which happen to be Chinese models. The speaker argues Anthropic's paternalistic restrictions are, paradoxically, accelerating the adoption of Chinese AI in sensitive research contexts. The absence of simple KYC controls, which Anthropic could implement immediately, is cited as the tell that safety is a pretext rather than the actual motivation.
The speaker draws a sharp policy distinction: regulate the output, not the model. Just as fertilizer sales require ID post-Oklahoma City and nucleic acid synthesis labs can screen orders, intervention should happen at the point where AI capabilities manifest as weapons—not by restricting who can access the underlying tools. Restricting model access denies economic and scientific upside while doing little to stop bad actors who can run open-source models locally.
On macroeconomics, CPI hit 4.2% YoY (highest since April 2023) and PPI came in at 6.5% YoY, with Fed rate-hike odds jumping to 49%. The speaker warns the Iran conflict risks an oil price spiral if Chinese strategic reserves enter the spot market, and argues the root cause of inflation remains excess government spending. A side discussion covers Bernie Sanders' proposed 50% one-time tax on major AI companies—dismissed as unconstitutional confiscation—and a counterproposal to reform Social Security into an equity-holding, account-based system.
The episode closes on the Los Angeles mayoral race, where the speaker contends the vote patterns—Nitia Ramen's post-election mail-in ballot surge concentrated in the Skid Row area, a guilty plea for vote-buying in that district, and laws making audits harder—amount to either illegal fraud or institutionalized legal fraud enabled by California's ballot-harvesting and universal mail-in laws. A federal DOJ audit is called for, and a California gubernatorial run by Steve Hilton is floated as a "break the glass" political remedy.
The episode covers Anthropic's controversial new model policies—silent downgrades, forced data retention, and restrictions on genomic and AI research—framed as evidence of regulatory capture. It also digs into suspicious ballot patterns in the LA mayoral race and rising inflation data amid the Iran conflict.
You can't just turn off AI — by stopping AI through political action and social behavior, you are fundamentally going to give someone else the advantage because the AI isn't going to go away
The lack of KYC implementation by Anthropic is the tell that they have a regulatory capture agenda.
The incremental cost of performing AI is excessive and large. That contrasts and compares to the incremental cost before AI which was zero.
We live in an autocracy in California where one political machine has shaped the laws so inexorably that whether it's a mayor or a state legislator or a state senator or it's the governor, it's a one-way door.
No evidence of fraud right now. Statistics don't look good. Donald J. Trump, you control the Department of Justice. Go investigate it.
The Ark Institute ingested all the world's genomic data they could get access to and built a genome language model
With a Kevin Worsh Fed, we could see north of five and a half, 6% overnight rates.
Meta really fumbled this with Llama.
The narrow piece they walked back is they are now saying that they will disclose when they downgrade you but they are still downgrading people when they decide that person should not receive the appropriate information
When Ben Thompson from Stratechery asked a very straightforward question about the relationship between cancer risk and GLP-1s, he was kicked out
Daario wrote a new blog post saying that transparency was no longer good enough and that we needed a new regulatory agency like an FAA or FDA to approve all models
CPI came in at 4.2% year-over-year, highest since April 2023.
Two years ago I bought 2,000 acres in Arizona with a partner, got it zoned and approved, allowed to build a 2 gigawatt data center
A gigawatt of compute now costs a hundred billion dollars
Most of the megawatts are still directed to the big models, and the overwhelming majority still goes to the big guys rather than open source
Calling for regulation of AI models would be a great way to scuttle and sandbag open-source models because they don't have the ability to be regulated
Anthropic released a model called Fable 5 on Tuesday and it tops nearly every benchmark
The tokens for Fable 5 cost twice as much as Opus 4.8, but it should use less tokens overall because it's theoretically that much better
In April Anthropic famously did not release Mythos because publicly it had some strong hacking capabilities
The CEO of Palo Alto Networks said Mythos was the real deal and that they used it to seal up some vulnerabilities inside their shop
While using Fable, Anthropic stores all the prompt data you enter for at least 30 days
If Fable 5 detects you're doing frontier AI research, they were downgrading you but not telling you, and this was buried deep in a 319 page document
Over the last couple of weeks, Anthropic has begun to restrict the ability to use the models to do genetic construct design work, citing bioweapon type risk
The best open source models today are Chinese
The American open source models are not as good as the Chinese open source models
Eight months ago, I said that Anthropic was engaged in a very sophisticated regulatory capture campaign based on fear-mongering
Even enterprise customers who had signed zero data retention agreements do not have a choice to opt out of the 30-day data retention for Fable/Mythos class models
Anthropic would degrade the product and nerf their models if it decided in Anthropic's sole discretion that you are not worthy of having access to that level of information, without telling you
Anthropic would rewrite your prompt in the background, give you a nerfed answer, not tell you what they were doing, still charge you for the product you thought you were getting, and never tell you that you were not getting Frontier model capability
90% of AI researchers and the most talented people in this industry, 80 90% of them think exactly like Daria.
Technology is fundamentally deterministic. Whatever is possible will be tried at least once.
Limiting the AI systems themselves is denying us both the economic opportunity, the job creation opportunity and also the world changing opportunity that these tools can enable.
Anthropic could implement KYC tomorrow, but it does not.
The international gene synthesis consortium of 2009 had all the major labs agree to develop and implement voluntary safeguards against misuse.
The models that are open sourced are already fully available, fully published, and anyone can access and use them.
The idea that you can quote regulate or downscale or turn off AI is not a realistic idea.
Bernie's proposal is a straightup confiscation of property and would be a terrible precedent.
The job loss apocalypse is not a story I believe in. We just had a blowout jobs report in May. 172,000 new jobs, over twice what economists were expecting.
Even software developer jobs are at a three-year high.
Daario actually has said we're going to have very high GDP but very high unemployment.
OpenAI and Anthropic are quote unquote public benefit corporations, meaning the corporations are not just supposed to maximize the value to shareholders but are supposed to do things that are in the public benefit.
The AI model downgraded the speaker from Fable 5 to Opus 4.8 after asking about fertilizer bomb regulations and nuclear bomb components.
The Social Security trust fund currently has one certificate in it. It is a certificate that is a very unique US Treasury certificate. The US government owes the Social Security trust fund $4 trillion.
There is no job loss with AI.
On the cost side of the equation, AI can be used to reduce humans doing things that cost money to some extent. The effect there, I would argue, is nominal.
The real opportunity with AI is on the revenue side where suddenly one engineer can do a hundred times or a thousand times what they used to be able to do.
Dario said specifically 50% job loss for entry-level knowledge workers in the next 1 to 5 years and he said that one year ago.
Sam has said that he basically I don't think he quantified, but he said that job loss was coming, but he said more recently that he was wrong because they're not seeing that in the numbers.
The public has been taught that AI means job loss and there's nothing in it for them.
last week they published this blog saying that recursive self-improvement could end the world. Therefore, we need a pause. What did they do the previous month? They hired Andre Carpathy to run recursive self-improvement and anthropic.
Ben Thompson made an excellent point that one of the reasons he thinks that they put out that blog post on the pause was to justify the anti-competitive step they took this week of depriving users of the ability to research AI, machine learning, and chip design using Fable.
The crazy thing about the internet, the reason why the Facebooks and the Googles massively overearned for decades is the marginal cost of production for a new user on the internet is effectively zero.
AI is completely different. There is a real cost for every marginal user. Everyone you stand up is taxing a GPU. Everyone you stand up needs electrons.
if you were going to go and take a piece of these companies, you have incredible leverage
Freeberg's idea is genius. That's why it'll never happen.
We should move to the same system that the Canadians, the Australians have proven works at scale.
SpaceX 100% anthropic 83% chance opening 48% chance they uh since our last episode disclosed that they have privately filed
the odds of a unicorn getting to decacorn was 8%. The odds of a decacorn getting to a senicorn was about double that was 13%. And then the odds of it going from senicorn to a trillion dollar market cap was 31%.
They train for free on all of our knowledge, but they're gatekeeping the output so that their competitors, which is a very large set of people, cannot use it.
Elon is describing an end state for humanity where the robots and AI are able to produce so much that people don't have to work if they don't want to
it's not Google who's the gatekeeping monopolist. It's now anthropic. They're not a monopolist yet. But if they get the right capture they're looking for, they will be either a monopolist or duopoly.
PPI came in at 6.5% year-over-year, highest since the end of 2022.
21% chance inflation hits 5% in 2026.
Chances of the Fed hiking this year at 49%. It was under 10% before the Iran war started.
European Central Bank just raised rates a quarter point on Thursday. This was their first rate hike since September 2023.
the core problem with wealth inequality in this country, the core problem with inflation in this country, the core problem, all of it roots back to excess government spending.
China has been smoothing the energy consumption globally worldwide and kept a damper on $200 a barrel oil.
The PPI number should be an alarm that we use to offramp this Iran situation sooner rather than later.
The PPI print was largely in line with expectations. That's why the market is actually up today.
NASDAQ up 2 and a half percent today.
In-person voting the day of the election in Los Angeles County for mayor: Spencer Pratt 35%, Karen Bass 29%, Nipia Ramen 26%.
Mail-in ballots received before election day: 38% for Bass, 28% for Pratt, 20% for Ramen.
All these ballots that arrive after election day: 37% for Ramen, 35% for Bass, 19% for Pratt.
Pratt's post-election mail-in ballots declined by one third while Nipia Ramen's increased by 80%.
The concentration of incremental votes that Nitia Ramen got came around the Skid Row area in Los Angeles.
California Assembly Bill 1921 legalized the practice of unlimited ballot harvesting in the state, passed around 2018.
California passed a law that made it permanent that every person registered in the state of California would get a ballot.
Anyone can register to vote in California without proving citizenship, using a gym membership card as an example.
California sends out ballots to every registered voter, something like 23 to 24 million people, but only 9 and a half million people vote.
When you sign your ballot, they run it through a machine and as long as the signature matches 40%, the machine says it's fine.
Gavin Newsome just signed a law making it virtually impossible to audit any of this stuff.
There's billions of dollars in fraud in California for things like Medicaid fraud, unemployment fraud, EBT fraud, hospice fraud, fraud related to the homeless industrial complex.
Brenda Lee Brown Armstrong pleaded guilty in 2026 of paying people on Skid Row to vote.
The election has been won by 43,000 votes.
The laws have been shaped this way to make legal what would otherwise be an illegality.
I don't think that there has been an election a major election that has been swung by election fraud
the Heritage Foundation has spent many many tens of millions of dollars on this
Lyndon Johnson only got elected to the Senate. He got a start in politics through ballot fraud.
we can't have a situation where someone is legally allowed to go out and collect a 100 ballots that they find sitting in the lobby of an apartment building
there's a ballot measure in California. If people in California want this, they can just vote for the ballot measure.
The Democratic party as a whole were election deniers about Trump winning in 2016. They concocted the whole Russia gate hoax to basically try and discredit his election.
When you look at the the difference between walk-in and mailin ballots... This data tells me that something is wrong.
the system itself has become a system of appointment, not a system of election.
The only thing we don't know is whether it was illegal fraud or legal fraud. Meaning they've created so many loopholes and so much corruption in the law that what they did is possibly legal.
I deny that Ramen won legitimately.
Newsome just signed a law making audits like that even harder.
I can tell you the statistical odds that this would have happened and it's one in a trillion.
Ramen registered one hour before the deadline and she's friends with Bass and that the whole point was to keep him from actually making it onto the general ballot.
ballot harvesting is legal, but it's sketchy
These powerful big tech companies are going to decide whether you're worthy, decide whether you're an AI have or have not, and then censor the output that you receive based on the criteria they determine when they profile you
If Anthropic has a deal with Novartis but not Eli Lilly, there's an impetus to shape how people get information about competing GLP-1 drugs
Anthropic has essentially shown their hand which is that they will increasingly take in prompts, evaluate the prompts and decide what to do with them before they generate output to you
A downstream scientist using the cloud APIs could trip one of these restrictions, and all of a sudden, you'll get cut off from a very important source of business differentiation for yourself
The restrictions that Anthropic and others are putting upon themselves and upon the industry is forcing a lot of companies to go and get open source Chinese models and run them
This is a trillion dollar company that's spending potentially billions of dollars on a regulatory capture agenda which is going to deprive you of access to those alternative models.
There'll be an AI monopoly or duopoly and they're going to decide along with some new government agency which will be a revolving door to their companies who has access to what capabilities.
Conservatives and libertarians are mortgaging their futures if they go along with this regulatory capture safetist agenda without really realizing that there's so much more to it at stake.
in the absence of power, this is all a moot conversation. And if you can't deliver power to hundreds and hundreds of megawatts and line of sight to gigawatts, it's all meaningless
If China somehow runs out of reserves and needs to go back into the spot market to buy an extra 3 million barrels a day, oil could reach between 150 and 200 a barrel.
The end result of California's political monopoly has yet to be unveiled because it's going to take another couple of decades for all of this rot to really destroy these places.
the fewer people actually believe that elections are real... that's when things get ugly and it needs to be fixed.
The media instead of trying to demand answers and demand accountability is trying to cover up and say nothing wrong happened here.
Companies need to start underwriting this next phase of AI, which is how do I have control, who am I allowing to learn off of all of this information, do I want to have single point of failure risk with respect to AI
There are three kinds of weapons enabled by AI: cyber weapons, physical weapons, and bioweapons. The capabilities that allow weaponization are the same capabilities that can cure cancer, make more food, and create software tools.
Regulation should be organized around the manifestation of AI tools in creating weapons, not around restricting access to the tools themselves.
Regulation should intervene at the output stage — e.g., at nucleic acid synthesis ordering — rather than at the access-to-model stage, similar to how fertilizer sales are regulated with ID requirements post-Oklahoma City.
We should change the system so that the Social Security trust fund can own equities. That should then turn into an account-based system so every individual has an account rather than a benefit.
It is not dissimilar to having the first set of transportation companies that run on the interstate highway. And if the interstate highways were built by the federal government writ large and now you had two companies that transported all the goods, a logical question at that time could have been how much of that should I own because you're riding on my rails?
There is a concept in insurance called adverse selection, which is if there is a moment to exploit something, the criminals will exploit it to the extent that it will become asymmetric.
There is one break the glass way... You must get somebody at the absolute top of the ticket elected. That person is Steve Hilton.
Both sides cheat. Both sides bend the rules. One side does it by limiting access. The other does it by providing too much access.
The only thing I can think of is her ground game is more sophisticated because they've been at it longer than Spencer Pratt and he didn't have a ground game
As a result of Anthropic's restrictions on genomic research use, Freeberg's company is likely going to end up needing to use open-source models and run them locally
The major AI labs along with lots of other signatories sent a letter in support of mandatory nucleic acid synthetic screening and recordkeeping.
Senator Sanders published an op-ed in the New York Times titled AI is a public resource. You should own half of it, announcing the American AI sovereign wealth fund act — a one-time 50% tax on stock of the largest AI companies including Open AI, Anthropic and XAI.
If you're a public benefit corporation, the board has to have a dual mandate balancing shareholders with the stated mission of the public benefit corporation.
I just had a review meeting with my product and engineering team two days ago and they're like, we want to add an extra 15 headcount to our engineering squads because we have all this opportunity to do stuff that we couldn't otherwise do.
It is time to vote all of these people out.
instantly on day one declare a state of emergency
There have been opportunities to pass laws that will make it illegal for people to vote if they are not citizens, if they are not properly registered, and if they don't show up with ID.
Trump has the ability with his hands-elected attorney generals to go after this... Go to this election. Use your powers. Audit it and find them.
“We're changing Fable 5 safeguards for Frontier LLM development to make them more visible”
“The foundation of AI is our collective and human intelligence. The books, the songs, the journalism, scientific research code essentially stolen by some of the wealthiest people in the world.”
“There is no job loss with AI.”
“The party told you to reject the evidence of your eyes and ears. It was their final most essential command.”
Shared with Earmark · earmark-ai.com