Perplexity Co-Founder Argues AI Safety Discourse Used to Concentrate Power; Turing Award Winner LeCun Agrees

Earlier this week, Perplexity and Databricks co-founder Andy Konwinski published an essay arguing that AI safety messaging is being weaponized to concentrate power rather than prevent harm. The essay followed Open Frontier, a working meeting of approximately 100 researchers convened by Konwinski through his Laude Institute nonprofit on June 30 in San Francisco.

Konwinski's argument centers on Anthropic's now-reversed decision to have Claude Fable 5 silently degrade responses for users suspected of training competing AIs, stating: "The problem isn't that Anthropic made a bad decision. The problem is that they assumed the decision was theirs to make." He proposes a research commons with frontier-scale compute as an alternative to centralized control. Turing Award winner Yann LeCun replied directly on X (July 3), stating the concentration of AI power is "by far the biggest danger of AI" and comparing the current moment to "medieval obscurantism akin to the Ottoman empire banning the use of the printing press for 200 years."