Overview

• OpenAI announced it may revise safety requirements if rival AI developers release high-risk systems without comparable safeguards.
• The updated Preparedness Framework introduces automated evaluations and risk thresholds for categorizing AI models by capability and potential harm.
• OpenAI's recent release of GPT-4.1 without a safety report has drawn criticism, raising concerns about reduced transparency and safety commitments.
• Former employees and critics warn that the company’s evolving approach could compromise safety standards to maintain competitiveness.
• CEO Sam Altman defended the changes, emphasizing the need to balance rigorous protections with user feedback and less restrictive model behavior.