David Sacks Tells OpenAI and Anthropic to Stop Pretending and Just Slow Down
David Sacks slams OpenAI and Anthropic for demanding AI regulation, arguing they can pace the frontier themselves without forming a cartel.
Policy · Source: Hacker News
What happened
David Sacks fired a massive warning shot at OpenAI and Anthropic. He publicly called out Sam Altman and Dario Amodei for their ongoing push to pace the frontier of artificial intelligence development. Sacks noted that these two companies currently hold a strict duopoly on frontier intelligence. They completely dominate market share, revenue growth, and model capability. They also claim their lead is widening rapidly due to recursive self-improvement. Sacks stated that if their unreleased models are truly as scary as they claim, they should just slow down voluntarily.
Sacks argued strongly that these companies do not need government permission to pause their work. They also do not need antitrust laws suspended so they can form a legal cartel. He accused them of using the threat of superintelligence to demand a preferred regulatory framework from the government. He pointed out that demanding regulations as the price for not building superintelligence looks like blackmail of the public and the political system. If they want to pace the frontier, they can simply agree not to build dangerous models.
The motivation behind this regulatory push is not pure altruism. Sacks pointed out massive product liability risks if their AI products enable a truly damaging cyberattack. He also called out METR, an AI evaluator group, as lacking independence due to its deep ties to Anthropic investors and staff. Sacks warned that if they do not simply slow down on their own, their push for regulation will look like a blatant bid for regulatory capture. He even suggested it could be an election-season psyop.
Key facts
- METR — An AI evaluator group that Sacks claims is intertwined with Anthropic investors and staff
- Hugging Face episode — A past event Sacks references to prove the market already punishes unpredictable models
- Fable 5 and Fable 5.1 — Models mentioned by a user that are currently blocking meaningful research work in the bio space
Why it matters
This public callout exposes the growing tension between open innovation and regulatory moats in the tech industry. Founders building on top of OpenAI and Anthropic need to deeply understand the financial and legal motives of their infrastructure providers. If the frontier labs trade raw power for predictability just to avoid product liability, the models you use will become increasingly restricted. We are already seeing this happen in real time. One user in the discussion noted that models like Fable 5 and Fable 5.1 actively block meaningful work in the biology space due to excessive safety guardrails.
The second-order effect here is a potential deep freeze on all emerging competitors. If OpenAI and Anthropic successfully push for a strict regulatory approval process, it will supersede standard product liability laws. This creates a massive and insurmountable barrier to entry for anyone else. Startups will not be able to afford the heavy compliance costs required by supposed independent evaluators like METR. The frontier labs will permanently lock in their duopoly while claiming they are just keeping the world safe from rogue artificial intelligence.
For builders
Prepare for heavily restricted frontier models
OpenAI and Anthropic are actively trading raw model power for predictability to avoid massive product liability from cyberattacks. Expect future model updates to feature much stricter guardrails that might break your current prompts or edge use cases. You will pay the price in lost model capability and flexibility.
Open source is your only viable hedge
Relying entirely on a duopoly pushing for regulatory capture is a massive platform risk for any startup. If new regulations mandate strict evaluations for all competitors, open-source models might be your only escape hatch. You need to start testing your core pipelines on alternative models today.
Compliance costs will crush new market entrants
If the proposed regulatory approval process becomes federal law, new AI startups will face insurmountable legal and operational costs. Evaluators tied to incumbents will police the entire market. Only the heavily funded players will survive this artificial compliance gauntlet.
My take
Sacks is completely right about this entire situation. OpenAI and Anthropic are clearly trying to pull up the ladder behind them. You absolutely do not need a government mandate to write safer code or slow down your own research. They want a regulatory moat to protect their massive duopoly and crush open-source competitors before they even reach the frontier. It is pure corporate strategy disguised as altruism.
Original reporting: Hacker News. This is my rewrite and opinion.