Perplexity Co-Founder: AI Safety Is an Excuse to Lock Down Frontier

Perplexity AI and Databricks co-founder Andy Konwinski thinks the AI safety conversation has a problem: It’s being used to concentrate power, not prevent harm. Earlier this week, he published an essay making his case, with Anthropic as the star witness.

The case he builds starts with a decision Anthropic reversed in 48 hours. When Anthropic launched Claude Fable 5 on June 9, a paragraph buried in its 319-page system card disclosed that the model would silently degrade its own responses for anyone it suspected of training a competing AI.

Researchers found it. The internet did not take it well.

Anthropic walked it back, but for Konwinski this makes no difference when analyzing the bigger picture. “The problem isn’t that Anthropic made a bad decision,” he wrote. “The problem is that they assumed the decision was theirs to make.”

His essay, titled “Concentration of power in AI is a risk, not a solution,” followed Open Frontier, a working meeting he convened through his nonprofit Laude Institute at San Francisco’s Exploratorium on June 30. About 100 researchers showed up.

UC Berkeley dean Jennifer Chayes, who runs the College of Computing, Data Science, and Society, told a funding panel that Berkeley researchers are “all building on Chinese models because we don’t have a Western open frontier model”—and that the safety messaging from OpenAI and Anthropic ahead of their IPOs amounted to a “very effective fear campaign.”

Konwinski’s argument is that centralizing access doesn’t neutralize risk; it creates a different one. AI is foundational infrastructure—in the same category as railroads, electricity, and the internet. Those technologies reorganized society around whoever controlled the underlying layer. The same is coming for AI. His alternative: a research commons with frontier-scale compute that lets top researchers reach the frontier without needing permission from a private lab to do it.

LeCun: It’s the Ottoman empire banning the printing press

Yann LeCun, Meta’s former chief scientist, replied to Konwinski’s essay on X with no ambiguity. “I’ve been disseminating a similar message for years,” he replied on Konwinski’s post. “The concentration of power in AI and the desire for control is by far the biggest danger of AI.”

He also had a historical comparison ready. “It’s a kind of medieval obscurantism akin to the Ottoman empire banning the use of the printing press for 200 years, in part to keep control of the dogma, but also to protect the corporation of the calligraphers and scribes,” LeCun wrote.

LeCun’s prediction for where this ends: “Infrastructure wants to be open. Foundation models are becoming an infrastructure and will inevitably become commoditized. Long term, the money is in the application layer.”

LeCun left Meta in late 2025 and launched AMI Labs in Paris with $1.03 billion in seed funding in March 2026—his own answer to the question. The company runs on world models and his JEPA architecture, plans to open-source its research, and has no commercial product expected for years.