The Public is Worried About AI. Silicon Valley Treats it as a PR Problem.
New interviews reveal how Silicon Valley power brokers view the AI backlash
Two recent interviews provide insight into how Silicon Valley power players are thinking about the public’s growing unease over AI.
Last week, David Sacks – Trump’s former White House AI and Crypto Czar – sat down with POLITICO’s Dasha Burns; this week, Chamath Palihapitiya – a prominent VC and media personality – appeared on Joe Rogan.
Both men are firmly among Silicon Valley’s aristocracy. Sacks was the founding COO of PayPal and later founded Yammer. In his brief stint at the White House earlier this year, he developed a reputation for forceful advocacy of laissez-faire AI regulation – and for getting under the skin of officials and Republican state leaders who pushed back. (The Renovator has previously covered his unseemly self-dealings while in the role). Palihapitiya was a senior Facebook executive in the late 2000s before ascending in the world of venture capital and earning the nickname “SPAC King” circa 2020 for his enthusiasm for Special Purpose Acquisition Companies.
Sacks and Palihapitiya co-host All-In, a business and tech podcast with over a million YouTube subscribers that frequently ranks as the #1 technology podcast in the US. Neither man is shy about his free-market views, his support for AI and Silicon Valley, or his wariness of government overreach. Sacks has called AGI a potential successor species. Palihapitiya has said “we are not going to fix governance; it may just be beyond repair.”
In each interview, both men were pressed on the American public’s widespread skepticism about AI. Their answers were telling:
Chamath Palihapitiya | The Joe Rogan Experience
Joe Rogan: How should we feel about a few companies making more and more and more? And then how do we feel about their ability to share that with a small amount of people? What is the expectation for everybody else?
Chamath: People need to see these corporate actors doing social good. In the Industrial Revolution, the leading lights of that era, Andrew Carnegie, Nelson Rockefeller, Jay Gould, J.P. Morgan, they sat together and they said, guys, this is going to benefit us, this Industrial Revolution. It may not benefit everybody. What is our responsibility? And they allocated tasks. Carnegie went and built libraries all throughout the country. Rockefeller built universities. Hospitals were built. And I think what happened is society was like, wow, these are living testaments to us doing well.
David Sacks | The Conversation with Dasha Burns (POLITICO)
Dasha Burns: Americans don’t really trust AI. What can and should companies and governments do to build that trust?
David Sacks: America is ahead in every category except one: optimism. Stanford did an international study of different populations’ views of AI and asked: do you think this will be more beneficial than harmful?” China was something like 83% AI optimistic. We’re at like 39%. I don’t blame the media exclusively, but the media is way too focused on the negatives of this technology as opposed to the positives. This has a huge impact on public discourse.
What are we to glean from these exchanges?
First, both men acknowledge that AI has a PR problem – the public isn’t as gung-ho about AI’s rapid rise as the Valley is. That’s notable in itself. The social media companies were famously slow to recognize shifting public sentiment, let alone confront it directly.
Second, their proposed solutions sidestep democratic oversight. Sacks wants better PR: reframe the narrative, highlight the positives, don’t let fear derail the race against China. Palihapitiya wants more philanthropy – tech barons doing visible social good, a la modern-day Carnegie libraries.
Let me pause here for a moment. These are just two excerpts from longer conversations. Elsewhere, Sacks acknowledges that targeted solutions to specific harms – child safety guardrails, ratepayer protection for data centers – are appropriate. Palihapitiya gestures toward tax reform, though not as a mechanism for democratic oversight.
But the clear thrust of both conversations is this: public concern is a liability to be managed, rather than a legitimate democratic signal worth responding to. Democratic participation and deliberation are an afterthought at best.
Then there’s the fact that Palihapitiya’s retelling of America’s Gilded Age is conveniently incomplete. What actually made society “okay” with the industrial transition wasn’t Carnegie’s libraries. It was coerced structural change: antitrust law, progressive taxation, labor organizing.
“Roosevelt led a democratic counterrevolution that reclaimed for democracy the ground that had been lost to capitalism,” historian H.W. Brands wrote of the era.
This worldview – that Silicon Valley doesn’t need to account for the views of the general public when developing and deploying technologies that transform our lives – has deep roots in the Valley’s founding mythology, from John Perry Barlow’s 1996 Declaration of the Independence of Cyberspace to Peter Thiel’s idiosyncratic lectures to Marc Andreessen’s manifestos. But Silicon Valley has never had as much sway in Washington as it does today, and has never wielded a technology this powerful.
Democracy as an End in Itself
In her book Justice by Means of Democracy (2023), Danielle Allen argues that democratic participation is not just a means to achieve some desired outcome, but an end in itself: justice requires everyone having a genuine voice in shaping the rules that govern their lives. To Sacks and Palihapitiya, democracy is useful when it produces the right outcomes, inconvenient when it doesn’t.
Nathalie Marechal, writing in Tech Policy Press this week, captures this perfectly: “forced to choose between accepting that democracy, rule of law and public-interest governance would necessarily result in reduced profit margins, or joining forces with a corrupt convicted felon with overt autocratic aspirations, the titans of the tech industry chose the latter.”
So what does democratic engagement with AI actually look like? It looks like residents of Valdosta, Georgia founding a local civic group and packing town halls to demand a say in the construction of a data center in their community. It looks like 1,000 Google DeepMind employees in London launching a bid to form what would be the world’s first union at a frontier AI lab. It looks like state and federal policymakers deliberating on the GUARD Act, the AI Lead Act, the TRAIN Act, New York’s Raise Act, and other AI-related bills.
None of this is of interest to Palihapitiya, Sacks, or most C-suite executives in the Valley. Democracy is slow and grinding. It is a speed bump to be flattened en route to AI accelerationism.
Fortunately, we still live in a constitutional democracy. Here, the people – and their elected representatives – get to shape the decisions that determine our collective future.



I appreciate this piece and the focus on democratic participation. One thing I’d add is that AI governance isn’t a single lever — it’s a four layer ecosystem, and each layer has different strengths and limits:
1. Standards bodies (NIST, ISO, IEEE)
They don’t regulate; they define how to measure safety, robustness, and transparency.
They create the legibility that everything else depends on.
2. Independent ethics & safety research
Academic and nonprofit groups that surface risks and frameworks.
They shape norms, even though they don’t have enforcement power.
3. Corporate governance
Internal guardrails like Microsoft’s Responsible AI Standard or Anthropic’s Constitutional AI.
Uneven, but real — and increasingly auditable.
4. Government bodies
The only layer with enforcement authority, but also the layer that uses AI for surveillance, national security, and intelligence.
The U.S. problem isn’t just “too little regulation.” It’s that we’re missing the connective middle layer that ties these pieces together — the layer other democracies treat as basic infrastructure.
If we want democratic oversight of AI, the path is likely a combination of standards, audits, procurement rules, and domain specific constraints (policing, immigration, critical infrastructure), rather than a single sweeping fix. It’s slower, but it’s how durable oversight is built.
Far more democratic participation is very, very warranted with the rollout of ai, among several other different dynamics currently at play. But that would require structurally re-decentralizing and returning to significant regional (states) and less-so-but-still-substantial local legal/regulatory variability, policy variability, re-decentralization of banking/finance (real decentralization, not BITCOIN!, so jurisdictional diffusion, credit policy variability, legally constructed pluralized capital structures, ect), deliberate redundancy, and local government fiscal primacy; among other things