David Sacks flagged a concrete risk for companies uploading proprietary data, trade secrets, and customer information to frontier AI labs: those same providers could later weaponize the transferred knowledge to launch competing products in verticals like design, code, or security. Chamath Palihapitiya amplified the point, noting the pattern already visible in sectors where AI can accelerate returns on capital employed.
🚩 David Sacks on Alex Karp’s CNBC ‘Crashout’
“Enterprises are at risk of transferring their knowledge, their know how, their trade secrets or customer data to these model providers who might eventually decide to compete with them. You can see that enterprises are waking up to this threat and they're not happy about it.”
Sacks pointed to Anthropic’s recent moves into Claude Design, Science, Security, Legal, and Code as evidence that model providers are expanding into customer domains, with one reported case involving a former partner blindsided by a competing launch. The concern is less about intent and more about the structural incentive created by access to enterprise data.
Both voices endorsed keeping data, model weights, and compute under enterprise control instead of relying on token-based services. Palantir’s recent emphasis on open-weight models via its Nvidia partnership is presented as one route, though the scale of adoption and long-term performance gaps remain unquantified in current statements.
Many users endorsed Sacks and Karp's warnings on data-leak risks when sharing with OpenAI and similar providers as insightful and truthful, while others replied with insults, accusations of fraud, or ideological attacks.
No Digg Deeper questions have been answered for this story yet.
I believe this 💯.
.@chamath says Palantir CEO Alex Karp is " an incredible, smart, brilliant guy" who "completely nailed it and called out on its face the huge risk" of giving your company's alpha to OpenAI and Anthropic:
"If you are a reasonable company, why are you not finding an independent way to access this intelligence in a way that doesn't leak your edge away?"
"To do so at this point now is kind of becoming derelict and irresponsible."
"Back then, you could be experimenting because you didn't know any better."
"Now, when you know all of these data points, to continue to make the same decision is really insanely dumb."
Via @theallinpod @DavidSacks @friedberg @Jason
Full Episode w/ @DavidSacks: @theallinpod
OK, look down, look waayy down, and you'll see millions of entrepreneurial peons.
That has been the stopper for me to integrate Claude into my one-man business.
The obvious upside is 360 degrees but the downside is that everything I enter into Claude or Grok or Perplexity or Chachi (I loke that), whatever - any of the AIs, I've now thrown my ideas into the ether.
In fact, if you're running the same prompt over five or sixAIs to get different perspectives, sometimes you could swear they're sharing information.
Apart from all the, I'm not talking about readily available Wikipedia stuff, but I'm talking about inside discussions that you might have with one.
That's my concern and, as I said, the concern of millions of little entrepreneurs that have not yet jumped into the AI pool.
@TheChiefNerd "Crash out" - Cringe Gen-Z slop
@TheChiefNerd Karp talks crazy and irrational because he is trying to avoid questions of fraud. He’s company is just an sql db that Claude can connect to. They have no AI themselves and ChatGPT release caught while company by surprise. Anthropic is palantir.
@TheChiefNerd He just confirmed none of this works on deployment without governance and AI OS system. Jensen just declared Service Now has the AI OS system. Them and Palentair own the governance and working with Microsoft that will be the loop closing.
@TheChiefNerd I totally agree with Karp here. And for some reason I think PTLR, NVDA and IREN need to work together to compete with Dario and Zuckerberg. Am I crazy for thinking this? Thoughts
@TheChiefNerd I’m pretty sure these “OPENAI” companies have stole my ideas… agree with @alexkarp
@TheChiefNerd Sacks is right. The model layer will commoditize. The distribution layer will not. Palantir understood early that the moat is not the algorithm. It is the deployment inside the customer's operations. Karp is defending a position that is already eroding.
@TheChiefNerd Enterprises are fine. They signed a non retention agreement. You guys are wasting lots of energy on a problem that’s already been solved.
@TheChiefNerd @WorldStrategist If that’s the concern, how are so many enterprises OK with using Chinese models?? Do they not understand the risks?
@TheChiefNerd It’s going to happen to Kashii. Microsoft is setting up a cloned app, “Arena”, that will push Kashii out within the year. It’s just so wrong.
@TheChiefNerd So what’s the difference between this and using professional artists work to train the models? Is it stealing or not?
@TheChiefNerd Enterprises spent a decade learning "trust but verify" with cloud vendors.
With model providers they skipped straight to "trust."
The fix isn't a better contract. It's verification:
http://InspectModel.com · http://CertifyModel.com · http://SafeModels.org
@TheChiefNerd This problem perfectly coincided with launch of GLM 5.2 which seems to be equally good at the coding and other tasks and 15x-20x cheaper. We are moving closer to the reality where models become commodities.
@TheChiefNerd Why the hell did it take this long for people to figure this out? They’re trying to create an omnipotent God, which all others will be subject to.
@TheChiefNerd that is what AWS does
@MichaelMaccaro5 @TheChiefNerd Palantir doesn’t compete w anyone lol- they are the only software that can make LLMs such as Claude, ChatGPT, grok- have any value lol
@TheChiefNerd Exactly! I have been saying that for ages.
@TheChiefNerd Brilliantly articulated…
