Quote · All-In with Chamath, Jason, Sacks & Friedberg
Anthropic's Digital God, Pope vs AI, Job Loss Narrative Flips, Open Source Crackdown Coming?
Where this was said
AI sovereignty, the next era of privacy, open-source crackdown coming?
At 51:05 · chapter starts 38:32
Jason Calacanis introduces a new concept: intelligence sovereignty. [1] — Jason Calacanis "Jason Calacanis introduces a new concept: 'intelligence sovereignty.' Privacy was about protecting your data from being read. Intelligence …" 38:15 Privacy used to mean protecting your data from being seen; intelligence sovereignty means preventing AI from analyzing your messages, emails, and photos to shape your worldview. He argues that open-source models running on Apple's M-series silicon (with 48GB–1TB of memory) are the only viable defense, and notes the paradox that Communist China is leading the open-weight movement while American companies push for centralization. [1] — Jason Calacanis "Jason Calacanis introduces a new concept: 'intelligence sovereignty.' Privacy was about protecting your data from being read. Intelligence …" 38:15 Chamath demonstrates that Fortune 1000 companies are already acting on this logic — demanding control planes that can hot-swap between frontier models to avoid vendor lock-in and political risk. [2] — David Sacks "Sacks sees a clear pattern: Anthropic repeatedly frames open-source AI as uniquely dangerous because guardrails can be removed. He believes…" 49:24 He shares a striking example: a Fortune 20 company that was tasked with generating $1 billion in AI OpEx savings instead burned $200 million on tokens in six months with minimal results. Meanwhile, a Polymarket post revealed that an AI consultant's client accidentally spent $500 million on Claude tokens in a single month. Sacks then delivers his open-source warning: [2] — David Sacks "Sacks sees a clear pattern: Anthropic repeatedly frames open-source AI as uniquely dangerous because guardrails can be removed. He believes…" 49:24 Anthropic's repeated rhetoric framing open-weight models as uniquely dangerous is deliberate predicate-building — inserting facts into the public record to justify a future ban. If America bans open-source AI, it puts itself on an island; the rest of the world runs on Chinese models. The result: a monopoly handed to two or three closed frontier labs, the precise centralization outcome the Pope feared.
Chamath explains that 80.90's Fortune 1000 clients refuse to lock into one AI provider, fearing both technology leapfrogs and ideological misalignment with a frontier lab's terms of service. They want a control plane that can hot-swap between OpenAI, Anthropic, or any open-weight model — because they see the model layer commoditizing fast.
A Fortune 20 CEO was asked for $1 billion in AI-generated OpEx savings; six months in, the company had spent $200 million on tokens with minimal measurable results.
Sacks sees a clear pattern: Anthropic repeatedly frames open-source AI as uniquely dangerous because guardrails can be removed. He believes this is deliberate predicate-building — putting facts in the public record to justify a future ban on open-weight models. A ban would shatter the competitive market and hand the monopoly to a handful of closed labs.
If the US bans open-source AI, the rest of the world will simply run on Chinese models — handing Beijing a decisive geopolitical advantage. Chamath highlights the paradox: America's adversary is championing open weights while American companies push for closed, regulated AI.
Elon Musk announced xAI rewrote its entire AI training complex in C, achieving an order-of-magnitude speed improvement and running on 220,000 GPUs — potentially collapsing training costs dramatically.
According to The Information, Anthropic is growing at roughly 10x year over year while OpenAI grows at ~3x — meaning Anthropic could command 90% market share within two years if rates hold.
