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

Perplexity AI and Databricks co-founder Andy Konwinski thinks the AI security dialog has an issue: It is getting used to pay attention energy, not stop hurt. Earlier this week, he printed an essay making his case, with Anthropic because the star witness.

The case he builds begins 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 mannequin would silently degrade its personal responses for anybody it suspected of coaching a competing AI.

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

Anthropic walked it again, however for Konwinski this makes no distinction when analyzing the larger image. “The issue is not that Anthropic made a foul choice,” he wrote. “The issue is that they assumed the choice was theirs to make.”

His essay, titled “Focus of energy in AI is a danger, not an answer,” adopted Open Frontier, a working assembly he convened by his nonprofit Laude Institute at San Francisco’s Exploratorium on June 30. About 100 researchers confirmed up.

UC Berkeley dean Jennifer Chayes, who runs the School of Computing, Information Science, and Society, informed a funding panel that Berkeley researchers are “all constructing on Chinese language fashions as a result of we do not have a Western open frontier mannequin”—and that the security messaging from OpenAI and Anthropic forward of their IPOs amounted to a “very efficient worry marketing campaign.”

Konwinski’s argument is that centralizing entry does not neutralize danger; it creates a distinct one. AI is foundational infrastructure—in the identical class as railroads, electrical energy, and the web. These applied sciences reorganized society round whoever managed the underlying layer. The identical is coming for AI. His different: a analysis commons with frontier-scale compute that lets high researchers attain the frontier without having permission from a non-public lab to do it.

LeCun: It is 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 have been disseminating the same message for years,” he replied on Konwinski’s submit. “The focus of energy in AI and the need for management is by far the most important hazard of AI.”

He additionally had a historic comparability prepared. “It is a type of medieval obscurantism akin to the Ottoman empire banning the usage of the printing press for 200 years, partly to maintain management of the dogma, but additionally to guard the company of the calligraphers and scribes,” LeCun wrote.

LeCun’s prediction for the place this ends: “Infrastructure desires to be open. Basis fashions have gotten an infrastructure and can inevitably turn into commoditized. Long run, the cash is within the utility layer.”

LeCun left Meta in late 2025 and launched AMI Labs in Paris with $1.03 billion in seed funding in March 2026—his personal reply to the query. The corporate runs on world fashions and his JEPA structure, plans to open-source its analysis, and has no business product anticipated for years.