Heritable editing of human embryos should remain prohibited until safety, consent across generations, and governance problems are substantially resolved. In other words, changes to an embryo's DNA could pass to later generations, so deployment would wait until safety and governance questions are much more settled.

Critical infrastructure should retain manual and offline backups even when full automation would be cheaper and more efficient. In other words, offline and manual backups reduce dependence on connected automated systems during cyberattack or system failure.

Routine police use of facial recognition in public spaces should be prohibited until accuracy, bias, and abuse risks are demonstrably controlled. In other words, facial-recognition systems compare images to stored templates and can produce false matches that vary across populations.

Large outdoor geoengineering experiments should require approval from a treaty-based international body and strong evidence that any harm can be reversed. In other words, countries participating in a formal international agreement would have to approve the experiment before it could alter shared climate systems.

Cities should deploy autonomous buses once independent evidence shows they are safer than human-driven fleets. In other words, autonomous buses operate without a human driver but may still use remote supervision and defined routes.

Gene-edited crops should be approved when they pass the same evidence-based safety standards as conventionally bred crops. In other words, gene editing changes selected DNA sequences and differs from adding genes from unrelated organisms.

Nuclear power should be expanded when modern designs can meet transparent safety and waste-management requirements. In other words, nuclear expansion can include existing reactor life extension, new large reactors, or smaller modular designs.

Clinicians should be allowed to use validated AI diagnostic systems with human oversight rather than wait for decades of additional experience. In other words, a diagnostic AI estimates conditions from clinical data; human oversight retains a professional responsible for the final decision.

Government should restrict automated systems that reproduce traditional gender and family assumptions, even if doing so slows deployment of useful AI. In other words, government would slow or limit an automated system when testing shows that it carries traditional assumptions about gender or family into important decisions.

Heritable gene editing should remain prohibited because altering future persons crosses a moral boundary even if it could prevent serious disease. In other words, heritable changes can affect descendants who cannot consent to the intervention.

Private firms should be free to deploy labor-saving automation once safety standards are met, without job-preservation quotas or public ownership requirements. In other words, job-preservation quotas require firms to retain a specified amount of human labor despite available automation.

Government should build and operate a national advanced-computing utility so access to powerful AI is not controlled by a few private firms. In other words, the government would own or operate large computing systems and allocate access under public rules instead of leaving the most powerful AI infrastructure to a few private companies.

Implantable brain-computer interfaces for elective cognitive enhancement should remain prohibited until long-term neurological and security risks are well understood. In other words, people should not yet receive implanted devices merely to improve memory or mental performance when the lasting risks remain uncertain.

Adults should be allowed to use nonheritable gene editing to enhance traits such as strength, endurance, or resistance to disease once safety is established. In other words, regulated biotechnology should be available not only to treat illness but also to improve healthy adult capabilities.