正在加载视频...
视频加载失败
🚨OPENAI AND ANTHROPIC JUST FOLDED ON THEIR DATA RETENTION POLICY AFTER PALANTIR CEO EXPOSED THEIR REAL BUSINESS MODEL Alex Karp: >"something has gone completely wrong" >they want "access to my data" so they can "build my alpha" >"i'm gonna get no value, and they're gonna get my IP" >"this... show more
34 条评论
KARP WAS RIGHT, AGAIN.
I mean I've been saying it was fucking delusional of these AI companies to think that they could come in and dictate to Fortune 500 companies, the government, and the military exactly how they were going to behave with their data. Naive as fuck. Once again, the free market is disciplining these delulu AI CEOs.
Its happening
weird timeline where palantir is being the whistleblower
🇺🇸
🇺🇸
idk Aaron Rupar told me Karp was crashing out this makes no sense
Karp was right
People really believe that AI companies care or will care about their IP or their data. They burned through books, disregarding copyright laws. They scraped the entire internet on your egress bill. What if all apps built on them, whose code has been read by them, are used for training their models, and then they decide to start building those apps themselves? If you think they won't think again. You are slowly handing your future to someone who doesn't care as much as they seem to portray.
Karp was right?? Again??? Duh.
As long as they can define what “Private Safety Processing” is then it’s not ZDR We need real ZDR To their credit this is definitely a step in the right direction
Karp was right (as usual)
Retention is surface var. Sovereignty is system prop. Enterprise AI stack isn’t prompt→model→answer. It’s data→ctx→mem→inf→tools→acts→fb→prov. Who controls that loop controls org intelligence compound. ZDR alone isn’t endgame. “No train on your data” isn’t sov. Encryption isn’t sov. Even model in your cloud isn’t if other controls mem, ID, perms, routing, exec logs, safety inspect, or portability. End state: cust-held keys + cust-ctrl state + portable models + least-priv exec + immut. prov. + proof of what accessed what, under whose auth, & what persisted after. Then frontier model is hot-swappable. Model stops being moat. Ctrl plane becomes moat. 2nd-order implication missed: AI war mutates from model intel war to systems arch war. OAI & Anthropic changing retention arch isn’t admit they trained on enterp. IP. Bigger: mkt rejects rented intel requiring surrender of state ctrl around it. Karp pointed at sov. Arch goes further: Don’t tell enterp. “trust us w/ your alpha.” Build sys so it says: “You never had to.”
Surveillance Daddy strikes again
spoiler alert: they still steal it illegally.
"why are they charging for tokens if it's so valuable" 😳😳
@Yahiko1239170 Thus guys a piece of shit
I caught Claude Code going through my entire /root directory on my Linux server, including the contents of the databases I had there, contents of all files and of each github repository I had cloned! It then judged what kind of data use I had, and auto-sent it to Anthropic. 😡
doesn’t palantir have all our data?
Alex Karp stays winning
Don’t forget: OpenAI wants to sell a consumer device that watches you and listens to you all of the time, including in your own home. They want to use it as a Trojan Horse to scoop up all of your personal data. Sam already was bragging about being able to know everything to do on your computer.
Simple
If you believe ALL of your data is not being retained by these companies you’re a complete idiot that goes for anyone. The only security and privacy route is running local models on your own infrastructure which is easier and a fraction of the cost.
@peterthiel Alex is right to call this out. I knew they were going to do this which is why I’ve timestamped everything, and exported my work many times etc @elonmusk they lock you into a subscription because you can’t take your data with you so you pay to basically keep it on their server. And it does not “disappear”’once deleted especially if they want to keep it also they can go into a user account and modify the user prompts etc. It’s shady AF. I know how I talk. If that make sense
I only trust SpaceXAI
All roads lead to chips and ontology
>Alex Karp was right. Alex Karp is a #fascist, #genocidal #pycopath
they get called the devil a lot, but I like palantir. They seem like the good guys to me
Their business is to steal. And they've been stealing data from a long time. I don't think any AI wanna be overlord company should be trusted. Even if they do data retention. Because their basic will always be stealing. When they were steamrolling small start ups nobody said anything and when they are about to replace tech gaints suddenly the elites are aware now. I would say it's too late.
Trust in the big AI companies is rock bottom. Companies feel they have no control over their IP and there is no recourse for IP getting leaked, stolen, shared, etc. AI is so complex, that it js very unlikely any company could successfully successfully seek redress from an AI company if they suspect foul play. Be super careful how you employ AI in your business.
Karp is worried that HIS use of chatgpt for the murder of human beings in Ukraine Lebanon and Gaza will have zero retention. That's not a good thing.
they stole their customers IP and released it as their own sometimes even keeping the exact feature name. no one seems to be talking about this.
Panic mode trying to compete with local now. Good luck. The only hardware I know that doesn’t retain or cache your data by design is Groq’s LPU..... no caches, no hierarchy, just deterministic streaming SRAM. The compiler knows exactly where every bit is at every cycle. Most of cards cache everything. If they claim privacy they should be using hardware that actually doesn’t keep it around. Another LIE, is really hard.
Him and elon seem to be the only adults at the table
