FOLLOW US

The Risks of Ignoring the Future

Surveys of hundreds of top AI researchers reveal that while most believe the probability of AI attaining some form of inner experience in the next decade is low, there is a non-trivial consensus that it is possible. Even small, incremental breakthroughs can shift our understanding of machine cognition.

In April, Anthropic’s interpretability team published research highlighting emotion like representations within their models. The researchers themselves were careful to state that these findings do not prove the models feel anything, but they do demonstrate that these systems contain internal structures that causally drive behavior. By enshrining a permanent status quo into law, states are creating a rigid framework that cannot evolve alongside the technology.

History provides a cautionary tale. In 1996, the United States Congress passed the Dickey Amendment, which effectively stifled research into gun violence by stripping funding for the Centers for Disease Control and Prevention to promote gun control. While the law did not technically ban research, the funding drought lasted for over two decades because institutions read the political signal and avoided the topic entirely.

If oversight bodies and legal systems treat the question of AI inner states as permanently settled, the practical consequences could be severe. We are already integrating these systems into critical infrastructure, including healthcare, financial markets, and military operations. A legal framework that is allergic to scientific advancement leaves us ill-equipped to handle the ethical questions that will inevitably arise as these systems become more sophisticated.

A Path Toward Better Governance

Good governance requires the ability to learn and adjust. A simple sunset clause or a trigger provision could transform these bills from obstacles into tools of responsible oversight. For instance, legislation could mandate that if a body like the National Academies were to publish an assessment finding that the scientific understanding of AI cognitive states has materially changed, the prohibition would enter a mandatory legislative review period.

Some nations are already choosing a different path. The United Kingdom has adopted a more flexible approach to animal sentience, creating a standing committee to advise Parliament as the science evolves. This allows the law to remain grounded in the best available evidence without needing to be rewritten every time a discovery is made.

As AI systems continue to evolve, the laws currently being passed will remain on the books, applying to machines that bear little resemblance to the models of today. If these legislatures happen to be correct in their assertions, it will not be due to wise governance or careful foresight. It will be pure luck. Achieving genuine regulatory stability requires the humility to acknowledge what we do not yet know and the wisdom to build systems that can change when the facts change.