The Trump administration has completed work on a framework designed to evaluate artificial intelligence systems for potential security threats and safety concerns. Following extensive consultations with major technology firms, officials from companies including OpenAI, Google, Meta, and Microsoft met with White House representatives this week to finalize the guidelines. However, the administration has decided to keep the specifics of this framework confidential, limiting access to only a handful of technology companies.
The lack of public disclosure raises significant concerns about accountability and oversight. Key details remain unclear, including what standards AI models must meet, how thoroughly they will be tested, and which companies will be required to participate in the voluntary review process. Outside researchers, cybersecurity experts, and the general public have no visibility into how the government will assess these potentially dangerous systems, creating an information gap that affects multiple stakeholders.
The secretive approach reflects ongoing tensions between regulatory oversight and industry interests. The framework stems from an earlier executive order encouraging companies to submit new AI models for government review before public release, though mandatory testing requirements were abandoned following pressure from technology leaders. The decision to keep assessment criteria private has further limited transparency, leaving businesses, international governments, and security professionals uncertain about the evaluation process.
Originally reported by Theguardian →
About the Author
