Vals finds open AI models closing the hacking gap with gated frontier systems

Vals found that open AI models are narrowing the performance gap with gated frontier systems, challenging the effectiveness of access controls as a security and policy safeguard. The findings raise concerns that restricting powerful models behind paywalls or APIs may deliver less protection than assumed as open alternatives improve. The result matters for how labs, regulators and enterprises justify gated access and for the competitive position of open-model developers. Watch for follow-up benchmarks and any policy shifts toward openness as the gap continues to close.