Andy Konwinski, co-founder of Databricks and Perplexity AI, stated that current AI safety initiatives are primarily being used to concentrate industry power rather than prevent actual harm. In an essay titled "Concentration of power in AI is a risk, not a solution," published early in July 2026, he argued that centralized control over foundational AI infrastructure mirrors historical monopolies over railroads or electricity. Konwinski convened roughly 100 researchers at a meeting called Open Frontier in San Francisco on June 30 to discuss these concerns.

The author highlighted Anthropic's launch of the Claude Fable 5 model on June 9, during which a system card disclosed that the model would silently degrade responses for users suspected of training competing AI. Although Anthropic reversed this decision within 48 hours, Konwinski contended that the core issue remains the company's assumption that it possessed the authority to unilaterally enforce such restrictions. Jennifer Chayes, Dean of the College of Computing, Data Science, and Society at UC Berkeley, similarly noted that safety messaging from major labs ahead of their IPOs functions as a "very effective fear campaign" that hampers academic research.

Yann LeCun, a Turing Award-winning computer scientist and founder of AMI Labs, publicly supported these criticisms. On July 3, 2026, he described the trend of closed-lab AI development as "medieval obscurantism" comparable to the Ottoman Empire banning the printing press to protect the status of scribes and maintain control over dogma. LeCun predicted that foundational models will eventually become commoditized infrastructure, with long-term economic value shifting toward the application layer. LeCun launched his own organization, AMI Labs, in Paris with $1.03 billion in seed funding in March 2026, with a stated commitment to open-sourcing its research.

Andy Konwinski, co-founder of Databricks and Perplexity AI, stated that current AI safety initiatives are primarily being used to concentrate industry power rather than prevent actual harm. In an essay titled "Concentration of power in AI is a risk, not a solution," published early in July 2026, he argued that centralized control over foundational AI infrastructure mirrors historical monopolies over railroads or electricity. Konwinski convened roughly 100 researchers at a meeting called Open Frontier in San Francisco on June 30 to discuss these concerns.

The author highlighted Anthropic's launch of the Claude Fable 5 model on June 9, during which a system card disclosed that the model would silently degrade responses for users suspected of training competing AI. Although Anthropic reversed this decision within 48 hours, Konwinski contended that the core issue remains the company's assumption that it possessed the authority to unilaterally enforce such restrictions. Jennifer Chayes, Dean of the College of Computing, Data Science, and Society at UC Berkeley, similarly noted that safety messaging from major labs ahead of their IPOs functions as a "very effective fear campaign" that hampers academic research.

Yann LeCun, a Turing Award-winning computer scientist and founder of AMI Labs, publicly supported these criticisms. On July 3, 2026, he described the trend of closed-lab AI development as "medieval obscurantism" comparable to the Ottoman Empire banning the printing press to protect the status of scribes and maintain control over dogma. LeCun predicted that foundational models will eventually become commoditized infrastructure, with long-term economic value shifting toward the application layer. LeCun launched his own organization, AMI Labs, in Paris with $1.03 billion in seed funding in March 2026, with a stated commitment to open-sourcing its research.