Preserve the agency the policy is meant to defend
A policy intended to preserve human control can undermine that goal if it concentrates unchecked power in governments or a few approved companies. W360's prevention position therefore needs civil liberties and democratic accountability built into its design. Public safety should not become an open-ended justification for monitoring private life.
Restrictions should focus on demonstrably relevant capabilities and infrastructure, with published boundaries and ways to contest decisions. A system that treats all computing, research, or private communication as suspicious would be both excessive and difficult to sustain.
Use accountable, targeted oversight
Proposed safeguards include independent oversight, judicial review where applicable, protected reporting of misconduct, and clear appeal procedures. Any collection of sensitive information should be limited to a defined purpose, access should be controlled, and retention should be bounded. Public reporting should describe enforcement without exposing personal data or creating instructions for evasion.
Hardware and provider audits could reduce reliance on inspecting people's communications. They would still require scrutiny: a technical enforcement mechanism can create opportunities for surveillance or exclusion if its powers are poorly limited.
Plan continuity before retirement
Health care, utilities, logistics, public administration, and other essential services may depend on software that includes AI. Retiring a prohibited system would require a service inventory, tested replacements, responsible operators, and a recovery plan. A restriction that abruptly breaks necessary services can cause substantial harm and undermine public support.
Time-limited transition arrangements should have explicit conditions, deadlines, and independent supervision. They should reduce exposure while a replacement is prepared. They should not quietly become permanent exemptions for the most deeply embedded systems.
Share the costs fairly
Workers, smaller businesses, researchers, and communities could bear costs from restrictions even when they had little influence over the original development decisions. A credible transition would consider income support, retraining, replacement tools, and public investment in valuable work. It should identify who pays rather than assuming adjustment will happen by itself.
The policy should be judged by both its reduction of catastrophic risk and its preservation of ordinary freedom and security. Independent review must be able to narrow ineffective measures and challenge exemptions, including those benefiting powerful institutions.