David Sacks on How Anthropic is Ironically Running Surveillance on Their Latest Models “This is the company that said that it was against government surveillance. They are now retaining for 30 days every prompt and every output you send to one of these Mythos class models.”

Full Episode w/ @DavidSacks: @theallinpod

We need a law requiring these models to be available for local offline use - these companies can’t be trusted at all.

🦆

Yep, this should be banned. People pay for their service. If prompts become a commodity, then Anthropic should pay back.

A beginner grabbing a super-sharp Japanese knife (like a high-end Yanagiba or Gyuto) is exactly like a regular user jumping straight into a Mythos-class model. The tool is incredibly capable, but it’s also unforgiving. One wrong move and you lose a fingertip — or in AI terms, you accidentally generate something dangerous, leak sensitive data, or trigger a safety incident. That’s why Anthropic (and most frontier labs) treat their strongest models differently

the retention point stands on its own, every prompt and output held for 30 days is a record that can be breached, subpoenaed, or repurposed, regardless of who's pointing it out

Sacks is just pissed off because @AnthropicAI is doing better than his heavy invested @OpenAI . Too bad, u lost. Deal with it but stop trying to tell the horse to wait for the donkey!

Yes, there is a major difference between a government nonconsensually mass surveilling their population versus a private company retaining data for training and safety purposes for 30 days.

Did he have a say in shutting down Fable or was he a complete bystander?

Sacks is on the administrations payroll to do exactly what he’s doing attempt to damage the credibility of any technology that does not fully submit to the administration anything related to FedGov & Tech coming out of this shill’s mouth should not be taken seriously

well no shit sacks… OAI does iI, too. it’s the greatest surveillance tool every created since zuck… except ppl talk to it like their dr and don’t put on a mask for the public ala fb… now you have to create anon profile to use the model and craft your queries wisely. pay to play per usual. race to the bottom for data surveillance throttle and ads for the plebs

This should suprise no one... There's a book written in the 60s that briefly describes a time when a small elite controlled AI, dictating the terms of life to exploit humanity by managing resources, governance, and military power effectively rendered people dependent on technology for survival and decision-making, leading to their enslavement and reducing them to mere cogs -- The book is Dune.

Anthropic has always had the build ai safely approach so not sure why people are surprised by any of this. Whether or not that is intentionally so they can mitigate competitors development, we will see

David Sacks @DavidSacks calls it out like it is! Anthropic/Misanthropic is complicated - "It's like your uncle who paid for you to go to college but molested you."

It’s hard for me to believe anyone would pay for that product. Misanthropic makes Google and Meta look trustworthy in comparison.

taken from Claude's android manifest. Note there are opt in caveats - but come on, man.

A disengenuous complaint, not least bc he's an early invester in Palentir, but also because underlying it is a strategy to eliminate other AI competition, & install an Open AI/Palentir hegemony- something we saw last night when the US admin. banned Fable 5..classic Thiel move

Private companies can log prompts for safety and users can walk away tomorrow. Governments can compel the data, punish based on it, and you cannot opt out. It’s the difference between a service you choose and power you cannot refuse.

This is the reason why you need verifiable ai

"Psycho-Eyes" Amodei and the Anthropic Goon Squad have competitors... I have a middle finger!

They’re kicking people off Fable and to Opus when asking questions on high risk topics. How is that a bad things

I love David Sacks, thank you for your service. Stay away from Anthropic

Does Sacks have financial interest in OpenAI winning? It certainly feels like it. Why would you put in VCs, angel investors and VC bros the be your "AI Advisors", it's totally corrupt with conflicts of interest.

Not a good precedent to set. If it is going to be done, it needs to be done under total transparency.

Gemini also keeps your data for 3 months. I own a Google Workplace account and Google will not let me delete my own data from my workplace that I pay for. Blew me away when I found out and I disabled Gemini though it still holds my data for 90 days. @grok could you confirm

No no no… this is not the most severe issue. The worst issue is, GIGO. AI is going to be a liar and shit due to the lies it is forced to take in. All these lies will make AI what? A liar. People who think they know better are creating omnipresent liars, worth nothing. Garbage.

How is retaining data any different from what Microsoft and Google do?

David Sacks pointed out Anthropic's irony: they keep every prompt and output for 30 days while claiming to oppose surveillance

How crazy for them is to implement the surveillance just before their IPO?

lol the irony of positioning yourself as privacy-focused while keeping 30 day logs on every interaction is hard to ignore

He is such a hater…at this point it’s just personal and not at all professional

$TAO -> 2k+ easily

what happens when you try to delete your prompt history?

THIS IS CRITICAL TO UNDERSTAND