Anthropic's Fable Backlash, Nationalizing AI, Inflation Heats Up & California’s Broken Elections
When AI companies implement mandatory surveillance and downgrade models based on your queries without disclosure, you lose access to powerful tools while competitors abroad face no such restrictions. The solution: demand transparency, support open-source alternatives, and recognize that restricting
1h 42mKey Takeaway
When AI companies implement mandatory surveillance and downgrade models based on your queries without disclosure, you lose access to powerful tools while competitors abroad face no such restrictions. The solution: demand transparency, support open-source alternatives, and recognize that restricting AI capabilities doesn't stop bad actors—it only handicaps legitimate users and gives adversaries an unfair advantage.
Episode Overview
The All-In podcast crew discusses Anthropic's release of their Fable 5 AI model, which implements controversial surveillance and content restrictions. The hosts debate the tension between AI safety concerns and the risk of regulatory capture, while exploring how these policies could disadvantage American businesses and drive users toward Chinese open-source models.
Key Insights
The Surveillance Trade-Off in Frontier AI Models
Anthropic's Fable 5 stores all user prompts and outputs for 30 days with no exceptions, even for enterprise customers with zero data retention agreements. The company builds profiles on users to determine what capabilities to unlock, and will downgrade model performance based on their assessment of your worthiness—sometimes without disclosure. This creates a new class of 'AI haves and have-nots.'
Regulatory Capture Through Fear-Mongering
While Anthropic implements strict controls on their models, CEO Dario Amodei simultaneously advocates for new government regulatory agencies to approve all AI models. This strategy effectively calls for restricting competitors (especially open-source alternatives) while positioning Anthropic as a gatekeeper. The approach resembles regulatory capture—using safety concerns to eliminate competition rather than genuinely addressing risks.
The China Open-Source Advantage
As American AI companies implement restrictions, they're inadvertently pushing businesses toward Chinese open-source models, which currently outperform American open-source alternatives. Companies needing uncensored AI capabilities for legitimate scientific work (like genomic research or materials science) are forced to adopt foreign models, giving China an economic and technological edge while American innovation is hamstrung by self-imposed limitations.
The Productivity Paradox: Same Tools, Different Uses
The capabilities that enable AI to create bioweapons, cyber weapons, or physical weapons are identical to those needed to cure cancer, develop better crops, create software tools, and empower entrepreneurs. Restricting access to the tools themselves prevents beneficial applications, while enforcement should instead focus on the manifestation of weapons and harmful uses—similar to how we regulate fertilizer sales after Oklahoma City rather than banning fertilizer entirely.
The Capital Requirements Death Spiral
Building AI infrastructure now requires staggering capital—approximately $100 billion per gigawatt of data center capacity, a 20x increase from initial estimates. This creates a massive barrier to entry for supporting open-source models, potentially forcing consolidation around a handful of closed providers who can then control access, pricing, and capabilities without meaningful competition.
Notable Quotes
"If you're a company, I think it's almost a non-starter. And the reason is because you could accidentally trip one of these things without even knowing it. A downstream scientist using the cloud APIs could trip it. A business executive inside your corporation could trip it. A person doing scientific molecular research could trip it. And all of a sudden, you'll get cut off from a very important source of business differentiation for yourself."
"It's just that the truth sucks when you actually take it and you eat it and you're like it's in your tummy and you're like no this is not good. I don't like this. And so there's the censorship risk on the one hand and then there's just the governance business risk for enterprises on the other. Both are not good."
"The restrictions that Anthropic and others are putting upon themselves and upon the industry is forcing a lot of companies to go and get open source Chinese models and run them. We're seeing this across the landscape with startups with large scale enterprises. Everyone's making that move."
"Technology is fundamentally deterministic. Whatever is possible will be tried at least once. And so we've already let this thing out of the box. So this idea that all of a sudden we can manage to get it back in and we're going to let a private citizen or a set of private citizens adjudicate what and to whom is insane."
"They would nerf their models if it decided in anthropic soul discretion that you are not worthy of having access to that level of information. So they're creating a new level of AI halves and have nots. And what they did is and they they this is there's a narrow piece here they walk back which is they said that when it came to things like machine learning AI research chip design research those types of areas they would kick you to a lesser model but not tell you that."
Action Items
-
1
Evaluate Your AI Provider's Terms of Service
Review your current AI provider's data retention policies, surveillance practices, and model downgrading terms. If you're using enterprise AI tools, verify whether you have zero data retention agreements and whether they're actually honored for new model releases. Consider whether you're comfortable with 30-day data retention of all your prompts, outputs, and context.
-
2
Explore Open-Source AI Alternatives
Research and test open-source AI models for your use cases, particularly for sensitive business applications or research work. Look into models like those from the Arc Institute for genomics, or evaluate Meta's Llama and other open-source alternatives. Consider the trade-offs between capability and control when selecting models for different applications.
-
3
Implement Your Own KYC and Governance Framework
If you're concerned about AI safety but want to maintain access to powerful capabilities, create your own Know Your Customer (KYC) framework and internal governance policies. Document who in your organization can access which AI tools, for what purposes, and with what oversight. This proactive approach demonstrates responsibility without surrendering control to external gatekeepers.
-
4
Diversify Your AI Infrastructure Strategy
Don't create a single point of failure by depending entirely on one AI provider. Develop a multi-model strategy that includes both closed and open-source options, and consider running some models locally or in your own infrastructure. This protects against vendor lock-in, unexpected policy changes, or access restrictions that could disrupt your business.