AI companies are increasingly restricting access to their most advanced models, raising concerns about transparency, accountability, and equitable access. OpenAI’s recently announced GPT-Rosalind, a powerful model for life sciences, is being made available only to select “trusted” users, reflecting a broader trend among tech firms to limit public access to high-capability AI systems.

While companies argue that restricted access helps manage risks associated with dual-use capabilities, such as cybersecurity and biological research, experts warn that allowing private entities to control access to such technologies raises governance challenges. Questions remain over who defines “legitimate” users and how global equity is ensured, particularly for researchers outside Western institutions.

The development also highlights tensions between safety and openness, as open-source models continue to evolve and may soon match proprietary systems. Experts are calling for stronger government oversight and clearer regulatory frameworks to balance innovation, risk management, and public accountability in the rapidly advancing AI landscape.