What People Are Getting Right and Wrong About AI Regulation Globally

By: Moish Peltz and Kyle Lawrence 

The global race to regulate artificial intelligence is accelerating, but policymakers are still struggling to keep pace with the technology itself. In the United States, Congress continues to debate broad AI governance frameworks while federal agencies and state governments pursue narrower enforcement actions and sector-specific rules. Meanwhile, regulators in Europe are already implementing sweeping frameworks like the EU AI Act and experimenting with technical standards around cybersecurity, interoperability, and infrastructure governance. 

That gap between policy ambition and technical understanding was a major focus of a recent episode of Block & Order, where Falcon Rappaport & Berkman attorneys Kyle Lawrence and Moish Peltz spoke with Wei Chen,, Chief Legal Officer of Infoblox and founder of the Atticus Project. Drawing on her work with regulators in both the United States and Europe, Chen argued that while governments increasingly recognize the need for AI governance, many still underestimate how quickly agentic systems are evolving and how difficult enforcement may become in practice.

Broad Agreement That AI Needs Governance 

One of Chen’s clearest observations was that policymakers, companies, and even frontier AI developers are no longer seriously debating whether AI requires oversight. Everyone agrees that frontier AI development needs clear guardrails, and the debate has increasingly shifted toward implementation. 

“I would say that what people get right is everybody agree that AI needs to be regulated somehow, or controlled,” Chen said during the episode. “They understand that there’s a need to provide governance, provide Trust. But they don’t understand AI behavior, especially agentic behavior, enough to identify the right things on which to put new guardrails.” 

That consensus represents a significant shift from earlier conversations around crypto and blockchain technology, where many policymakers questioned whether regulation was necessary at all. In AI, even the largest companies developing frontier models increasingly acknowledge that some form of governance framework will be unavoidable. Chen pointed specifically to growing concern around advanced models and agentic systems capable of operating independently online. While she stopped short of endorsing every public warning issued by major AI labs, she noted that increasingly powerful systems create legitimate governance concerns that regulators cannot simply ignore. 

At the same time, Chen suggested that governments are often attempting to regulate technologies they do not yet fully understand. That disconnect becomes especially pronounced when policymakers attempt to govern autonomous agents, machine-to-machine interactions, or infrastructure layers that have no obvious analogue in traditional Internet regulation. 

The Technical Understanding Gap 

Much of the conversation focused on what Chen sees as one of the biggest weaknesses in current AI policymaking: a lack of technical fluency among regulators and lawmakers. 

When discussing her conversations with regulators about agent discovery systems and agent-to-agent communication, Chen described encountering repeated confusion about why those systems matter in the first place. As AI systems become more autonomous, agents will 

increasingly need to locate, authenticate, and communicate directly with other agents online in order to complete tasks efficiently. That could include everything from coordinating software actions and retrieving data to negotiating transactions or interacting with outside infrastructure without human intervention. 

“When I was talking with regulators about agent discovery, it’s kind of foreign to them,” she explained. “They don’t understand why agent needs to discover another agent.” 

That problem becomes more significant as AI systems move beyond simple chatbot interfaces and toward increasingly autonomous infrastructure. Chen outlined what she described as a “three phase of agentic transformation,” beginning with today’s AI systems navigating human-designed interfaces and eventually evolving into direct agent-to-agent communication at machine speed. 

Her concern is that regulators are still focused primarily on the first phase, while infrastructure decisions are already shaping what future systems will look like. “We all have heard about MCP and A2A,” Chen said, referring to emerging agent communication protocols. But “everybody is still in this agent-to-UI place.” 

That lag matters because infrastructure standards tend to become difficult to change once adoption reaches scale. Chen repeatedly returned to the importance of interoperability and open systems, particularly as large technology companies compete to become dominant AI platforms. 

Governance Frameworks Need Technical Enforcement 

But even intelligently designed regulations are ineffective if there are no systems in place to enforce them, Chen explained. 

“I think when lawyers today, or governance professionals, think about AI governance and privacy rules and all of those things, you have to think about the code,” Chen said. “Otherwise [a law is] just a piece of paper sitting somewhere.” 

She pointed to areas like DNS infrastructure, software development kits, identity layers, and machine-readable policy enforcement as examples of where governance ultimately needs to function in practice. Without those implementation mechanisms, organizations may adopt formal AI policies that are effectively impossible to operationalize. 

Governments already struggle to consistently enforce existing rules around spam, cyberattacks, and digital fraud. As AI tools become more powerful and more accessible, those enforcement challenges may become significantly harder. Chen compared the situation to physical-world deterrence systems, noting that society generally imposes meaningful penalties for certain forms of offline misconduct while enforcement online often remains fragmented and inconsistent. That imbalance could become increasingly problematic as AI systems lower the barriers to sophisticated cyberattacks, impersonation schemes, and autonomous malicious activity. 

Regulation Will Continue Chasing the Technology 

One of the most practical takeaways from the episode was Chen’s acknowledgment that even experts deeply embedded in the industry struggle to keep up with the pace of AI development. 

“AI is moving so fast at a dizzying pace,” she said. “Even I couldn’t keep up.” 

That reality creates a difficult challenge for regulators attempting to build durable governance systems around technologies that evolve on a near-monthly basis. It also creates practical compliance questions for businesses trying to evaluate risk while the underlying legal framework remains unsettled. 

As AI regulation continues developing globally, companies operating in this space will increasingly need advisors who understand both the legal frameworks and the technical infrastructure underlying these systems. Falcon Rappaport & Berkman’s Artificial Intelligence Practice Group, including Kyle Lawrence and Moish Peltz, works with companies navigating AI governance, infrastructure regulation, privacy issues, and evolving compliance obligations. 

Businesses evaluating AI deployment, agentic systems, or cross-border governance risks should consider engaging counsel early as the regulatory environment continues to evolve.