Price:

AI & TECH

Sacks warns AI rules enable state censorship

Monday, July 20, 2026 · from 4 podcasts
All-InThe a16z ShowAI Daily BriefThe Daily
  • David Sacks warns AI safety mandates are pretext for state-mandated ideological filtering and surveillance.
  • Major AI labs push pre-approval regimes to crush startups and open-source rivals.
  • Export restrictions backfire, pushing allies to Chinese AI stacks like DeepSeek and Huawei.

David Sacks sees the real AI threat not in superintelligence, but in state control. On The a16z Show, he argued that algorithmic discrimination laws in states like California and Colorado force developers to embed DEI filters into models - making AI sanitize truth to fit political agendas. When Google’s Gemini rewrites history, it’s not a glitch. It’s policy in action.

Sacks calls this 'Orwellian AI' - a system where the government leverages safety rhetoric to mandate ideological compliance. He cites the Biden administration’s executive order, which included about 20 pages of DEI language, as a blueprint for embedding state values into AI. This trust-and-safety infrastructure, ported from social media, becomes a surveillance tool under civilian control.

The same day, Nathaniel Whittemore reported on The AI Daily Brief that Trump’s AI executive order was pulled twice before signing - after Sacks intervened. The final version stripped out mandatory pre-approval, reduced model review from 90 to 30 days, and made safety testing voluntary. The NSA leads testing, but the system lacks teeth. Critics like Steve Bannon see it as a 'foot in the door' for future mandates.

"The biggest risk isn't Terminator - it's 1984. It's AI that lies, distorts, and rewrites history to serve a political agenda."

- David Sacks, The a16z Show

A week later, on All-In, Sacks doubled down. He endorsed Demis Hassabis’s proposal for a FINRA-like self-regulatory organization - but only if it excludes speech controls and includes open-source players. He warned that Anthropic’s push for an 'FAA for AI' is really a bid for regulatory capture. Their strategy, he claims, is to lobby for patchwork state rules that ratchet up restrictions and lock out smaller players.

Meanwhile, U.S. export controls are backfiring. By blocking GPU sales and model access to allies like Saudi Arabia and the UAE, Washington is driving them toward Chinese alternatives. The launch of DeepSeek and Huawei’s Cloud Matrix disproved the idea that China was behind. Sacks argues the U.S. is losing the diffusion race - shrinking its ecosystem while subsidizing Beijing’s.

"If you hand the entire rest of the world to China’s tech stack, you’ve already lost."

- David Sacks, All-In with Chamath, Jason, Sacks & Friedberg

#Regulation#AI & Tech#Politics
Demis HassabisNathaniel WhittemoreAnthropicUAEDeepSeekChinaGoogleSaudi ArabiaNational Security Agency

Source Intelligence

- Deep dive into what was said in the episodes

All-In with Chamath, Jason, Sacks & Friedberg
All-In with Chamath, Jason, Sacks & Friedberg

Can the AI Industry Regulate Itself? Stripe Wants PayPal, China Catches Up, NY Bans Datacenters • Jul 18

  • Demis Hassabis proposes a FINRA-like U.S.-led international AI standards body: industry-funded, federally overseen, run by independent experts.
  • David Sacks outlines five criteria for the proposed AI SRO: broad industry representation including startups and open source, focus only on frontier models, limit scope to catastrophic cyber and CBRN risks, start voluntary, and be a substitute for a new government agency.
  • David Sacks warns Anthropic's regulatory strategy aims for patchwork state rules ratcheting up restrictions, citing Politico's 'Inside Anthropic's State-by-State Plan' article.
  • David Friedberg notes token spend at enterprises grew 21 times over the last year and CFOs lack control over AI API costs, which could cause earnings misses.
  • David Friedberg lists token prices: $56 per million tokens for Fable, $26 for Seoul and Quad 4-8, $1.50 for GROC and Zuck's model, $0.50 for Chinese models.
  • David Friedberg cites OpenAI's blog post 'PRC linked influence operations are targeting AI debates in the U.S.', suggesting foreign influence campaigns shape U.S. data center opposition.
  • David Friedberg describes Calico and Revel Pharma using AlphaFold and directed evolution to create a novel enzyme that degrades CML, reversing skin glycation from over 70 to 31 years old.
Also from this episode: (6)

Startups (2)

  • David Sacks says PayPal's stagnation stems from eBay's 2002 acquisition and corporate mindset purging founding DNA, creating the PayPal diaspora.
  • Jason Calacanis notes a new M&A wave for mature digital businesses: Ryan Cohen bid for eBay, Stripe-Block-Advent bid for PayPal, and Bending Spoons' roll-up of Vimeo, Evernote, and Eventbrite.

Business (2)

  • David Friedberg details a PayPal bid structure: Stripe and Block contribute $17 billion equity, Advent provides cash, and operators likely take over post-acquisition.
  • David Sacks argues defining the market as Visa/MasterCard duopoly makes a Stripe-Block-PayPal merger pro-competitive, not anti-competitive.

Energy (1)

  • David Sacks notes a PG&E auction secured only 156 megawatts for 7-8 gigawatts needed, highlighting a severe U.S. electricity shortage.

AI Infrastructure (1)

  • David Friedberg details Elon Musk's 'behind the meter' strategy using mobile turbines to bypass clean air permitting for data center power.
#Banking#Regulation#Macro#Startups
The a16z Show
The a16z Show

Replay 2025: David Sacks on AI, Crypto, and America's Technology Future • Jul 16

  • David Sacks says the European approach to AI leadership focuses on defining regulations in Brussels, not technological innovation.
  • During the Biden years, the SEC pursued crypto via 'regulation through enforcement,' prosecuting companies without clear rules.
  • Sacks says the Trump administration's mandate for crypto is pro-regulation - to provide clarity so the industry can comply and the U.S. can compete.
  • President Trump declared in a Nashville speech he would make the U.S. the crypto capital of the planet and fire SEC Chairman Gary Gensler.
  • Sacks claims crypto founders faced personal debanking - they couldn't open personal bank accounts, a form of extreme censorship.
  • A crypto summit at the White House under Trump was a milestone; an attendee said a year earlier they'd have expected jail over an invitation.
  • Anthropic engaged in a regulatory capture strategy, aiming to require government pre-approvals for new AI models, according to David Sacks.
  • Anthropic co-founder Jack Clark admitted making people afraid of AI was part of their strategy to push for regulation like SB 53.
  • Sacks argues Silicon Valley's success is built on permissionless innovation, unlike heavily regulated industries like pharma or defense.
  • The Biden administration's last-week 'Biden diffusion rule' required government licensing for every GPU sale on Earth.
  • Over 1,200 bills to regulate AI are in state legislatures; 25% are in California, New York, Colorado, and Illinois.
  • Algorithmic discrimination laws in Colorado, Illinois, and California make model developers liable for outputs with disparate impact on protected groups.
  • Colorado law defines people without English proficiency as a protected group, making model outputs about 'illegal aliens' potentially illegal.
  • The Biden executive order on AI contained about 20 pages of DEI language, aiming to promote DEI values in AI models.
  • Sacks sees the biggest AI risk as Orwellian control - AI that lies, distorts answers, and rewrites history to serve a political agenda.
  • Sam Altman predicts automated AI researchers by 2028; David Sacks is skeptical, arguing a virtual AI researcher requires AGI, not creates it.
  • Andrej Karpathy now says AGI is at least a decade away, pulling back from the imminent AGI narrative.
  • Sacks describes the current AI landscape as polytheistic - many specialized models making progress - not a single monolithic god-like intelligence.
  • He argues AI is 'middle-to-middle,' needing human prompting and validation, while humans are 'end-to-end,' setting objectives.
  • Sacks says AI is hyper-democratizing, with about 600 million users rapidly heading toward 5 billion across consumer products.
  • The best AI models are in consumer products like ChatGPT or Grok; you cannot spend more money to access a better AI.
  • The AI market is hyper-competitive with five major model companies leapfrogging each other, not consolidating into a monopoly.
  • Sacks argues open source AI is synonymous with software freedom, allowing users to run models on their own hardware.
  • Half the global data center market is on-prem, with enterprises and governments running their own data centers instead of using big clouds.
  • The irony is the best open-source AI models are Chinese, a reversal of expectations where the U.S. promotes open and China promotes closed.
  • President Trump's July 23rd AI policy speech declared the U.S. must win the AI race, emphasizing innovation, infrastructure, and exports.
  • A single federal AI standard is needed to avoid a burdensome patchwork of 50 different state regulatory regimes.
  • Sacks argues the Biden administration's export restrictions drove Gulf states like Saudi Arabia and UAE into buying Chinese chips and models.
  • AI doomerism is replacing climate doomerism as the left's central organizing catastrophe to justify controlling the economy and information.
  • The effective altruist movement pivoted from pandemics to existential AI risk after Sam Bankman-Fried's fraud, pushing for consolidated control.
  • Biden staffers told Mark Andressen they would ban open-source AI and anoint 2-3 winners, comparing AI regulation to Cold War physics bans.
  • Top Biden AI employees went to Anthropic immediately after the administration ended.
  • The launch of DeepSeek and Huawei's Cloud Matrix disproved the narrative that China was far behind and would copy U.S. regulations.
  • Regulators predicted frontier models trained on 10-25 flops were too risky; all frontier models now use that level of compute.
  • The Genius Act passed the House with about 300 votes and 78 Democrats; the Clarity Act needs 60 Senate votes under filibuster rules.
  • The Clarity Act passed the Senate with 68 votes, including 18 Democrats.
  • David Sacks says the Democratic Party's future appears to be woke socialism, with all major figures endorsing Mandani in New York.
  • San Francisco has a weak mayor system; the Board of Supervisors and left-wing judges constrain the mayor's power.
Also from this episode: (8)

Open Source (2)

  • Open source is a catch-up strategy; it attracts non-aligned developers and commoditizes software, complementing China's hardware manufacturing.
  • Reflection, founded by former Google DeepMind engineers, is a promising open-source initiative in the West.

Models (1)

  • The U.S. leads in closed AI models but is behind China in open-source models.

Macro (1)

  • The U.S. large national market is a fundamental competitive advantage, unlike Europe's fragmented pre-EU landscape.

AI Infrastructure (2)

  • Energy is the basis for the AI boom; shedding 40 hours of peak grid load to backup generators could free up 80 gigawatts of power.
  • Huawei's Cloud Matrix technology networks 384 Ascend chips to compensate for individual chip inferiority versus Nvidia.

Energy (2)

  • The immediate bottleneck for data center power is a 2-3 year backlog for gas turbines, not a shortage of natural gas.
  • Nuclear power will take 5-10 years to scale, so gas is the short-term solution.
#Regulation#Safety#China#Open Source#Energy
The AI Daily Brief: Artificial Intelligence News and Analysis
The AI Daily Brief: Artificial Intelligence News and Analysis

Nathaniel Whittemore

AI Optimism vs. AI Pessimism • Jul 14

  • Nathaniel Whittemore states the Trump administration's AI executive order was signed after being pulled twice, with the final version making safety testing voluntary and reducing the pre-release review period from 90 to 30 days.
  • Whittemore says David Sacks intervened to stop the initial signing, and the NSA is assigned primary responsibility for testing under the order with support from cyber and defense agencies.
  • Dean Ball argues the executive order is a win for safety advocates and tees up a future licensing regime, while Steve Bannon predicts mandatory testing will be implemented within months.
  • Nathaniel Whittemore notes Anthropic expanded its Project Glasswing, adding 150 new partners across 15 countries in sectors like energy, healthcare, and communications to test its Mythos model.
  • Whittemore cites The Information's report that Mythos testers are spending millions on tokens, with Anthropic subsidizing costs, and firms are planning budgets to build strategies around the model.
  • Whittemore highlights OpenAI's report showing Codex growth, noting it has 5 million weekly active users and non-technical knowledge workers are adopting it three times faster than developers.
  • Nathaniel Whittemore says OpenAI identifies three frictions in knowledge work: finding inputs across sprawling systems, information coordination costs, and approvals and verifications.
  • Whittemore states 72% of Codex knowledge workers produce artifacts weekly, 41% do research, 27% perform data analysis, and 15% implement business workflows, with half now running parallel tasks.
  • Nathaniel Whittemore describes Codex's new features: Annotations for precise document interaction, role-specific plugins bundling 62 apps and 110 skills, and Sites for turning artifacts into shareable web apps.
Also from this episode: (4)

Chips (1)

  • Nathaniel Whittemore reports SK Hynix plans to double memory chip capacity by the decade's end to address AI-driven shortages, with Chairman Chey Tae-won stating the shortage could last until 2030.

Enterprise (3)

  • Nathaniel Whittemore notes Uber has instituted a $1,500 monthly token spending cap for employees, signaling cost management as a key vector for the next wave of enterprise AI.
  • Whittemore says Microsoft announced seven new AI models, including the 1-trillion parameter MAI Thinking 1, positioning them as part of a cost optimization strategy for enterprise adoption.
  • Nathaniel Whittemore cites Mustafa Suleyman stating that when tuned for McKinsey's tasks, MAI models outperformed GPT-5.5 on quality while being ten times lower on cost.
#Regulation#Big Tech#Models#Agents#Enterprise
The Daily
The Daily

Why the Cease-Fire With Iran Keeps Crumbling • Jul 14

  • The US-Iran ceasefire agreement has collapsed due to ambiguous wording in Paragraph 5, which states Iran will make best efforts for safe passage through the Strait of Hormuz, but does not define which channels must be open.
  • David Sanger argues the agreement was rushed and poorly constructed by inexperienced negotiators (VP Vance, Jared Kushner, Steve Wickoff), lacking input from diplomats familiar with Iran’s history of exploiting loopholes in agreements.
  • Iran interprets the agreement as requiring them to open only one channel in the Strait of Hormuz (the one closest to their coast), while the US expects all three channels to be open.
  • The Iranian government’s motivation for closing the Strait is not purely economic; it is rooted in pride and nationalism, demonstrating resistance to US pressure and asserting sovereignty.
  • Iran possesses two key leverages against the US: the ability to build a nuclear weapon quickly and the capacity to disrupt global energy markets by blocking the Strait.
  • The US has initiated a blockade of Iranian ports to cut off oil revenue, returning to a stance of economic pressure after the ceasefire failed.
  • David Sanger notes that the Trump administration fundamentally misjudged Iran’s revolutionary ideology, treating the conflict as a rational economic deal rather than a 47-year ideological opposition.
  • Current negotiations are stalled, with no direct talks between senior officials; communication occurs via intermediaries like Qatar or Pakistan, unlike the Obama-era direct engagements.
  • Sanger states the US is entering a phase of repeating failed strategies, such as blockades, without a clear plan for a comprehensive nuclear agreement or conflict resolution.
  • President Trump’s public suggestion of imposing a 20% security charge on cargo in the region contradicts the stated US goal of restoring free passage through the Strait.
  • A federal immigration agent shot and killed a person in Maine during an operation targeting an individual believed to be in the country illegally.
  • Governor Henry McMaster appointed Darlene Graham Nordone to fill Senator Lindsey Graham’s Senate seat following his death.
Also from this episode: (1)

War (1)

  • The conflict has shifted from potential full-scale war to a prolonged, low-intensity engagement, akin to a counter-terrorism operation, with the US Navy likely remaining in the Strait indefinitely.
#War#Diplomacy#Middle East#Trade

Related Stories