Academic experts Sheldon H. Jacobson and Daniel Solow argued Wednesday that artificial intelligence regulation should focus on technical capabilities and risks rather than human-like characteristics. The professors stated that while advanced systems like ChatGPT and Claude can simulate human conversation, they possess a different cognitive architecture that requires specialized governance.
The calls for a technical approach to regulation follow the rapid development of generative AI systems. Jacobson, a computer science professor at the University of Illinois Urbana-Champaign, and Solow, an operations professor at Case Western Reserve University, noted that AI capabilities are now evolving on timescales of months rather than generations, complicating efforts to predict future system behavior.
The authors detailed specific risks that necessitate oversight, including the ability of AI to breach cybersecurity defenses and solve complex problems independently. They noted that autonomous actions by AI do not require human motives or consciousness to cause harm, but emphasized that motives remain human when people use these systems for personal gain. They argued that effective regulation must address a system’s specific access, autonomy, and potential for harm.
A person using these technologies would notice changes in how AI is deployed in their day-to-day life, particularly in the safeguards applied to digital services and cybersecurity. The scale of this impact is tied to how quickly the technology evolves; the authors noted that AI is currently solving problems that humans have historically struggled to solve, which could alter professional roles in data-driven industries. The timeline for these shifts is compressed, as AI capabilities can emerge within months.
The policy approach suggested by the authors sets a precedent for moving away from anthropomorphizing technology in legal frameworks. If adopted by regulators, this would shift the focus of future policies toward technical audits of a system's capabilities rather than debates over machine intent. What happens next depends on how government agencies incorporate these technical risk assessments into formal rules, though no specific legislative deadlines or vote dates were reported in the source.