Both sides, in good faith

The strongest fair case each way — we don't pick a winner.

The case for

Supporters of the administration's decision to keep the framework closed-door argue that testing methodologies and benchmarks for evaluating AI systems are sensitive by nature, and that publishing them in full could allow developers to game evaluations rather than build genuinely safer systems. They point out that such frameworks are often working documents, refined through ongoing technical consultation with agencies and industry, and that premature disclosure risks locking in immature standards or inviting politicised lobbying before the details are settled. On this view, a measure of confidentiality during development is standard practice in government procurement and safety testing, and reflects prudence rather than evasion.

The case against

Critics, including much of the tech policy community, argue that a framework determining how government tests and approves AI models before public release is a matter of significant public interest, and that keeping it hidden undermines accountability and trust. They contend that companies need visibility into benchmarks to prepare for compliance, that independent experts need to scrutinise whether the testing is rigorous and free from political influence, and that months of anticipated openness followed by a closed-door reversal fuels suspicion about how standards were set and for whose benefit. From this perspective, transparency is not a bureaucratic nicety but essential to ensuring the process is credible and fair to all developers.