Major artificial intelligence companies are increasingly pushing for independent reviews and safety evaluations of their flagship models before deployment. A central point of contention involves striking a balance between rigorous public transparency and protecting sensitive proprietary code and trade secrets. Critics and civil society groups argue that voluntary, industry-backed reviews without mandatory public disclosures risk functioning as corporate public relations rather than accountability. Policymakers are grappling with how to mandate standardized safety evaluations while ensuring that disclosure rules do not inadvertently hand blueprints to malicious actors. The ongoing debate highlights the friction between commercial pressures to stay competitive and the public’s demand for verifiable safety guarantees.

Full report : Why the AI lab’s advocate for third-party oversight of advanced models yet stop them from reviews.