The rapid development of increasingly capable AI systems is creating a regulatory problem that voluntary industry safeguards may not be sufficient to solve. The underlying concern is that AI capabilities are advancing faster than governments and institutions are developing mechanisms to evaluate, constrain and respond to dangerous behavior. Wright’s broader work on AI has consistently focused on the possibility that increasingly autonomous systems could create risks that are difficult to reverse once they become sufficiently powerful.
A central issue is the emergence of agentic AI. AI systems are moving beyond answering questions toward writing and executing code, using tools, interacting with external systems and coordinating subtasks. Wright has specifically highlighted the transition into this agentic phase and the growing importance of systems such as Claude Code. As autonomy increases, the potential consequences of a failure become much larger: a model that produces a bad answer is one thing, while an agent with access to software, networks or business systems can potentially act on that mistake.
The regulatory debate therefore isn't simply about restricting AI innovation. It is about establishing mechanisms that can determine when a system becomes dangerous, require meaningful safety testing and create accountability when companies deploy systems with significant real-world capabilities. This is particularly important because developers have incentives to compete aggressively and release increasingly capable models. Relying entirely on companies to decide how much capability is safe could create a race in which safety becomes secondary to reaching the next technical milestone.
The broader argument is that AI safety should be treated more like other high-consequence technologies than like ordinary consumer software. Regulation does not necessarily mean stopping development; it can mean requiring testing, monitoring, transparency and safeguards proportional to a system's capabilities. The urgency comes from the possibility that once highly autonomous AI systems are widely deployed, governments may have considerably less ability to impose effective controls. Wright's position fits into the larger 2026 debate over whether humanity should wait for clearer evidence of catastrophic risk—or establish stronger safety rules before the evidence arrives at potentially enormous cost.
