In recent years, major AI labs have been proactively establishing internal audit systems aimed at improving model safety. However, critical analyses have gained attention within the industry, pointing out that this approach lacks a fundamental process for ensuring safety.
While many AI labs are deploying specialized teams within their development processes, transparency and independence have become central points of discussion. Critics note that the current frameworks rely heavily on a company's own proprietary standards, often neglecting objective verification by third-party institutions and rigorous safety checks prior to market deployment.
As the societal impact of AI models continues to grow, it is suggested that self-contained audits alone may be insufficient to properly control risks. Moving beyond internal monitoring, the establishment of highly transparent auditing processes involving external experts and rigorous "front-door" check systems before introducing products to the market has become essential. Discussions surrounding the formulation of trusted AI development standards are expected to accelerate further moving forward.