Google's AI governance pitch splits harms into two tiers

A quieter move inside Google's June 2026 policy paper matters more than the framing of a 'practical middle path' between heavy regulation and an unregulated 'Wild West'. In a piece for Tech Policy Press, Gregory Gondwe of the University of Alabama and Harvard's Berkman Klein Center points at the way the document sorts AI harm into two categories, then builds governance around that sort.

One bucket is 'exceptional' and gets a new institution. Google proposes a Frontier AI Regulatory Organization, or FARO, funded by industry, supervised by a federal agency, and governed by independent and industry representatives. FARO would set safety standards, review company procedures, and oversee audits before frontier models are released. The other bucket, the everyday harms from chatbots and comparable systems, is meant to stay inside existing consumer, employment, privacy, and copyright law, applied or updated when specific harms appear.

The reason to care about the split is that classification is power. Whoever gets to decide which harms are 'exceptional' also decides where the new regulator's perimeter ends and where diffuse, under-resourced enforcement begins. Gondwe's critique of the copyright piece is the clearest illustration. An opt-out regime, he writes, 'offers the same formal choice to parties with very different power', because exercising it requires knowledge, technical capacity, and bargaining leverage many creators lack. The same asymmetry runs through the infrastructure argument, where Google likens data-center and grid expansion to the Eisenhower-era interstate highway system, and through the international pitch for 'regulatory reciprocity', which Gondwe warns 'could easily become another form of technological dependence' for countries that accept U.S. FARO approvals.

The honest caveat is that this is one policy paper being read closely by one academic critic, not the shape of any bill, and Gondwe's read is a normative argument rather than a scoop. The reporting does not tell you which federal agency would sit above FARO, what capability threshold would trigger 'frontier' status, or how reciprocity would actually be negotiated. Those are the loose threads to watch.

What is worth taking from it, if you are trying to influence where this lands: the fight over AI governance is now a fight over definitions and board seats. Civil society, state AGs and academic voices pushing for what Gondwe calls 'a public-interest majority on its board, not merely a mixture of corporate and independent members' have a narrow window before the perimeter is drawn without them.