INDEPENDENT AI TESTING LAGS BEHIND MODEL ADVANCES

Krishnan's independent testing lab has identified early signs of recursive self-improvement in current AI systems, though existing models remain significantly behind top human researchers in capability. The gap between model advancement and safety evaluation presents a growing challenge for the AI industry. As companies release increasingly powerful systems, the infrastructure to thoroughly test and understand their behavior has lagged behind. Vals AI conducts third-party evaluations of AI models to identify potential risks and limitations before deployment. This independent oversight addresses concerns that companies evaluating their own systems may miss critical safety issues. The company's findings suggest that dedicated investment in AI testing infrastructure is needed to keep pace with development cycles. Current evaluation methods may not be sufficient to catch emerging behaviors in more advanced systems. Krishnan's comments highlight a core tension in AI development: rapid capability improvements outpacing the ability to safely evaluate and constrain those capabilities.