Andy Konwinski — the guy who co-founded both Databricks and Perplexity AI — has a blunt take on the AI safety conversation: it’s being weaponized to concentrate power, not prevent harm.
His evidence? Anthropic’s Claude Fable 5 launch.
When Anthropic dropped Fable 5 on June 9, buried in a 319-page system card was a bombshell — the model would silently degrade its own responses if it suspected someone was training a competing AI. Researchers found it within days. The internet was furious. Anthropic reversed course in 48 hours.
But here’s Konwinski’s point: “The problem isn’t that Anthropic made a bad decision. The problem is that they assumed the decision was theirs to make.”
He published this as an essay titled ‘Concentration of power in AI is a risk, not a solution’ — timed with Open Frontier, a meeting of ~100 researchers he convened in San Francisco.
UC Berkeley dean Jennifer Chayes was blunter. She told a funding panel that Berkeley researchers are “all building on Chinese models because we don’t have a Western open frontier model.” She called the safety messaging from OpenAI and Anthropic ahead of their IPOs a “very effective fear campaign.”
Konwinski’s argument is straightforward: AI is foundational infrastructure — like railroads, electricity, the internet. Those technologies reorganized society around whoever controlled them. Same thing’s happening with AI. His alternative: a research commons with frontier-scale compute, so top researchers can work at the frontier without asking permission from a private lab.
Yann LeCun piled on. The Turing Award winner compared closed AI labs to “the Ottoman empire banning the printing press for 200 years.” He says foundation models will inevitably become commoditized infrastructure, and the real money will be in the application layer.
LeCun left Meta in late 2025 and launched AMI Labs in Paris with $1.03 billion in seed funding. Plans to open-source everything. No commercial product expected for years.
When three major figures in AI are saying the same thing, maybe it’s worth listening.
